The board approved the spend. The IT team checked the boxes.  The licences are assigned. Yet, months later, you are staring at a dashboard that reveals a grim truth –   your people are not actually using the tool. 

According to the Worklytics benchmark data, organisations struggle with an active user rate below 45 per cent. If you are paying £19.32 per user, per month for Microsoft 365 Copilot and more than half of your seats are gathering digital dust, you are effectively burning £10.63 per licence, every single month. 

This is not a technical failure but rather an adoption failure. The problem is not the software, it is the ‘people stuff’. Organisations hit this wall because they treat AI like a minor update instead of a fundamental rewiring of the workday. 

The classic “Honeymoon Plateau” pattern

Most rollouts follow a predictable and disappointing arc. 

It begins with a spike of curiosity known as the ‘Honeymoon Phase’, where employees ask Copilot to write a poem or summarise a single long email. But once the novelty of the ‘magic trick’ wears off, usage craters. Users revert to their old, manual habits because they have not been shown how to weave AI into the fabric of their actual jobs. 

This plateau is the result of a massive disconnect between the capability of the tool and the daily reality of the user. As the research indicates, “Adopting AI is as much a people and process challenge as a technology one.” If you do not align the software to specific workflows, it remains a nice-to-have toy rather than a mission critical engine. 

The five real reasons your adoption stalled

If your usage numbers are flatlining, you are likely suffering from one of these five strategic friction points. 

1. Weak prompts equal weak outputs

Trust dies in the prompt box. When a user enters a shallow prompt like “Summarise this project,” they get a generic, useless output. They conclude the tool “doesn’t work” and never come back. High maturity users utilise Prompt Depth by writing over 100 words to provide goals, context, and specific sources. 

Example of a deep prompt:

Act as an expert executive assistant and data analyst. Search for and retrieve all internal emails, calendar invites, and Teams chat threads or channel messages referencing Project X that were sent or received over the past 14 days. 

From this data, compile a comprehensive, pre-meeting briefing document structured specifically for our review session tomorrow. The briefing must synthesise the retrieved information into four distinct sections: 

1. Schedule status: Detail any timeline slips, completed milestones, or upcoming critical dependencies. 

2. Cost issues: Highlight any budget variances, unexpected expenses, or cost concerns raised by the team. 

3. Open risks: Identify newly introduced or escalating project risks that require immediate mitigation. 

4. Deferred decisions: List all technical or strategic decisions that were explicitly put on hold for senior leadership review. 

Ensure the final output is concise, action-oriented, and free of redundant conversational chatter, highlighting who raised each key point where applicable. 

2. The workflow gap

You have given them a tool without a map. Adoption only sticks when the tool is integrated into people’s specific daily routines. A workflow is only ideal for Copilot if it meets three conditions:

• It has repetitive steps 

• The data lives in Microsoft 365 

• It relies heavily on summarising or analysing 

3. Leadership as an optional extra 

Behavioural shifts are top down. If executives are still manually drafting reports and ignoring AI generated insights in meetings, they are signalling that Copilot is optional. Leadership must be the primary practitioners, not just the sponsors. 

4. The social proof deficit

AI literacy is socially contagious. Stalled organisations often lack the Proof mechanism where colleagues share ‘aha!’ moments. Without seeing a colleague save two hours on a report using a specific prompt, the average employee remains immune from the benefits of Copilot. 

5. The single training rollout

Launch day is not the finish line, it is the starting block. Knowledge evaporates without continuous reinforcement. If your training ended on Day 1, your adoption likely ended by Day 30. 

Benchmarking ‘good’, and waht high maturity adoption looks like

To move from a laggard to a leader, you need to track impact and deep integration, not just basic enablement. By using the native Microsoft 365 Copilot usage reports, you can compare your stats versus what “‘good’ looks like: 

• Active Users Rate (85 per cent target): Which means nearly every enabled (licensed) user is taking at least one intentional action using an AI-powered capability within the 28-day measurement period. 

• High Usage Intensity & Consistent Usage (11+ actions target): Track stats such as Usage Intensity and Average Prompts Submitted per User to ensure users are engaging frequently ( taking 11 or more actions per month, for example) and consistently returning week after week, indicating genuine AI habit formation rather than one-off experimentation. 

• Copilot Assisted Hours (Targeting high estimated hours saved): Multiply your employees’ total Copilot actions (like drafting emails or summarising meetings) by time-saving multipliers (e.g., estimating 6 minutes saved per summary action) to estimate the total hours of manual work eliminated. 

• Active Agent Users (High organisational penetration target): Measure Active Agent Users and Feature-level Actions. This tracks how many unique users are interacting with custom-built Copilot agents or moving beyond basic chat to use advanced features (like Excel data analysis or creating PowerPoint presentations), indicating they are transforming specific business workflows. 

Lessons from the UK Government deployment

The UK government did not just turn Copilot on; they designed a rigorous evaluation framework to measure its true value. They deployed the tool to roughly 20,000 employees across multiple departments and agencies. The results were highly encouraging, with an average of 26 minutes saved per user per day, which equates to roughly 13 working days saved per year. Additionally, 82 per cent of participants stated they did not want to return to working without the tool. 

The Department for Work and Pensions conducted a deeper six-month trial of 3,549 users, proving that the tool delivers massive value when applied to routine, structured tasks. 

Their adoption blueprint focused on three key areas. 

• Task matching: Identifying specific everyday administrative tasks where the tool could immediately eliminate manual effort, such as retrieving complex policy information and drafting routine emails. 

• Time reinvestment: Guiding employees to consciously reinvest their recovered hours into higher value public service tasks and strategic project planning. 

• Phased departmental enablement: Moving away from a broad launch to focus on targeted department where teams could share learning experiences directly. 

They didn’t just buy licences. They built a framework around adoption. 

Looking ahead

IT leadership in the AI era is no longer about platform management; it is about behavioural and cultural leadership. Licences are a commodity. The ability to fundamentally change how your people think and work is your only sustainable competitive advantage. 

Failing to reach high maturity adoption is not just a budgetary oversight but a strategic risk. While your organisation remains tethered to manual, low velocity processes, your competitors are using AI to reinvent their entire value chain. 

Are you ready to move from implementation to real scale? 

Join the webinar 

“Delivering Real Value from Microsoft Copilot from Implementation to Scale with Cloud Direct.” Learn the specific behavioural shifts required to move your organisation into the top quartile of AI maturity. 

For the last two years, the AI conversation has been dominated by copilots, chat assistants, and increasingly powerful language models. The demos have been impressive and the pilots have been promising, but for many organisations, the reality has felt underwhelming. 

At best, AI has often behaved like a highly sophisticated Q&A machine. Useful? Absolutely. Transformational? Not always. 

The challenge has never really been intelligence. Today’s AI models are already incredibly capable. The real challenge is context. 

Microsoft Build 2026 brought that shift into focus. Microsoft’s vision, which they’re calling Microsoft IQ, moves beyond model capability and towards context. Rather than focusing solely on model capability, it introduces a unified context layer that helps AI understand how work actually happens, what business data means, and what it should or shouldn’t have access to. 

The result is a meaningful shift in how organisations should think about AI adoption. 

Why are AI projects struggling to deliver value? 

We’re all familiar with the fact that AI can summarise meetings, draft emails, analyse documents and answer questions. However, when asked to support complex business processes, connect information across teams, or make informed recommendations, it frequently falls short. 

This is because AI often lacks three critical forms of context: 

• An understanding of how work happens across the organisation 

• An understanding of what business data actually means 

• An understanding of organisational rules, policies and permissions 

Without these foundations, AI remains disconnected from the reality of the business. Microsoft IQ is designed to close that gap. 

What is Microsoft IQ? 

Microsoft IQ is Microsoft’s vision for providing enterprise AI with organisational context. 

Rather than treating AI as a standalone assistant, it creates a foundation that helps AI understand the people, processes, data and governance structures that sit behind the business. 

Right now, Microsoft IQ has three context layers: 

• Work IQ 

• Fabric IQ 

• Foundry IQ 

Who benefits most from Microsoft IQ? 

The organisations likely to benefit most from Microsoft IQ are those looking to move beyond basic AI use cases. 

The real opportunity lies in helping AI support decision-making, operational processes and end-to-end workflows. 

Teams that will benefit most include: 

• Sales teams managing complex customer relationships 

• Project teams coordinating multiple stakeholders 

• Operational teams handling large volumes of information 

• Business analysts working with organisational data 

• Leadership teams making strategic decisions 

In reality, almost every knowledge worker can benefit when AI understands the context surrounding their work. 

How Work IQ helps AI understand work 

One of the most important components of Microsoft IQ is Work IQ. Think of it as the organisational memory of your business. 

Work IQ helps AI understand how work actually happens by connecting information from meetings, conversations, emails, documents and collaboration tools. 

Instead of simply locating information, Work IQ helps AI understand: 

• Who owns a project 

• Which stakeholders are involved 

• Where decisions were made 

• What actions are outstanding 

• How teams collaborate 

Rather than acting like a search engine, AI starts behaving more like a knowledgeable colleague who understands the wider context surrounding a task. 

How Fabric IQ gives AI business context 

Fabric IQ helps AI understand business data. Traditional AI can access data, but often lacks an understanding of what that data represents. 

For example, an AI model might see a figure of £500,000 without understanding whether that number represents revenue, cost, profit, budget or forecast. 

Fabric IQ adds business meaning to data. 

It connects AI to: 

• Business metrics 

• Semantic models 

• Organisational definitions 

• Data relationships 

• Performance indicators 

This enables AI to understand not only what the data says, but why it matters. Instead of simply reporting figures, AI can interpret performance against business objectives and provide more meaningful insights. 

How Foundry IQ keeps AI secure and governed 

The third pillar of Microsoft IQ is Foundry IQ, which focuses on governance, security and organisational knowledge. 

Enterprise AI must operate within clear boundaries. It needs to understand what information exists, who can access it, and how it should behave. 

Foundry IQ helps ground AI in: 

• Internal policies 

• Governance frameworks 

• Security controls 

• Compliance requirements 

• Organisational knowledge 

Without this layer, AI introduces risk. With it, organisations can deploy AI with greater confidence and trust. 

How Work IQ, Fabric IQ and Foundry IQ work together 

The real value of Microsoft IQ emerges when these context layers operate together. Imagine a scenario where regional sales performance suddenly drops below target. A traditional AI assistant might flag the decline and produce a report, but an AI agent powered by Microsoft IQ goes much further. 

Work IQ brings in the human context to surface the teams, projects and conversations connected to the issue. 

Fabric IQ layers in business context, analysing performance data against sales targets and broader objectives. 

At the same time, Foundry IQ ensures everything operates within the right guardrails, limiting access to authorised data and enforcing organisational policies. 

The result is not just a summary, but a recommendation grounded in real context. Instead of presenting disconnected insights, AI combines human, business and governance context into a single, coherent view that’s far more useful for decision-making. 

When should organisations start preparing?

The short and simple answer is now. 

Not because Microsoft IQ is the latest technology trend, but because context is rapidly becoming the foundation of successful AI adoption. 

 Most organisations already have access to powerful AI tools. The next challenge is getting consistent value from them at scale. 

That requires organisations to focus on: 

• Data quality 

• Information architecture 

• Governance 

• Security permissions 

• Knowledge management 

• Clear business processes 

The organisations that invest in these foundations today will be best positioned to take advantage of the next generation of AI capabilities. 

Is your organisation ready for context-aware AI? 

As you move beyond experimentation and towards AI at scale, success will increasingly depend on the quality of your data, governance and business context. 

If you’re exploring Microsoft Copilot, AI agents, Microsoft Fabric, or preparing your organisation for the next wave of AI innovation, getting those foundations right is what makes the difference. 

We spend a lot of time with organisations who’ve already rolled out Copilot or started experimenting with AI, but aren’t seeing the value they expected. 

If you’re in that position, we can help you work through it. Contact us below.  

By Leon Godwin, Cloud Evangelist

During our recent webinar, Mastering Security and Governance in Microsoft Fabric, I was joined by Microsoft’s Rana Kamel to unpack one of the biggest tensions we’re seeing amongst IT teams right now.

The session covered a lot of ground and generated a huge amount of interest, which means there were some great questions that we just didn’t get to in time. So I’ve pulled together your questions and answered them here to give a bit more clarity.

How does security actually work across OneLake, shortcuts, and SQL endpoints? 

This came up in lots of different ways, but the core concern was the same: 

Does data stay secure as it moves across Fabric? 

The short answer is yes, but only if it’s configured properly. 

• OneLake enforces security through identity. Every request is validated through Microsoft Entra ID, which helps prevent cross-workspace or cross-tenant leakage. 

• Shortcuts respect the original source permissions through pass-through identity. Users still need access to the underlying data, not just the shortcut itself. 

• Sensitivity labels and governance rules flow with the data, so encryption and export controls are maintained. 

Where people get caught out is SQL endpoints. 

• To apply OneLake security properly, SQL Analytics Endpoints must run in User Identity mode. 

• If you use Delegated mode, OneLake security is bypassed and you need to manage permissions manually. 

So nothing “magically bypasses” security, but misconfiguration can create gaps. 

How should we approach access control at scale?

Another big theme was how to manage access when different users need different views of the same data. The key is layering your controls rather than relying on one mechanism: 

• Workspace roles control who can access environments 

• Row-level and column-level security control what data users can see 

• Microsoft Purview adds governance through classification, lineage, and policy enforcement 

At scale, this becomes less about individual permissions and more about design: 

• Use governed, certified datasets 

• Apply consistent patterns across Lakehouses 

• Centralise policies where possible 

On ABAC specifically, the practical equivalent today comes from combining identity, roles, and Purview policies rather than relying on a single ABAC model. 

Do security rules carry through the medallion architecture? 

Security does not automatically carry through from Bronze to Silver to Gold. 

• Each layer is effectively new data, created through transformation 

• That means security has to be reapplied or redesigned at each stage  

In practice, many organisations: 

• Apply stricter controls at ingestion in Bronze 

• Refine access in Silver 

• Enforce business-ready security models in Gold 

The key takeaway: do not assume inheritance. Design for it. 

When should we move to a shared Lakehouse model? 

Many organisations start with a tenant-per-workspace model for isolation, which is a safe approach but it doesn’t scale forever. The recommendation is to move to a shared Lakehouse only when: 

• Your pipelines are fully automated and auditable 

• You can enforce row-level security reliably 

• You have clear tenant partitioning in place 

This is less about a fixed point in time and more about maturity. 

How does Fabric prevent data leakage in multi-tenant environments? 

Fabric provides protection at the platform level: 

• Compute workloads run in containerised Spark sessions that do not share memory 

• Every interaction is validated through identity to prevent cross-tenant access  

That said, platform security alone is not enough. Poor data design or missing governance can still introduce risk, especially when introducing AI tools. 

What should we put in place now if we want to introduce AI and Copilot later? 

AI doesn’t create new security issues, it exposes the ones you already have. The best approach is to design with governance from day one: 

• Implement data classification and DLP policies 

• Ensure end-to-end data lineage for auditability 

• Certify trusted data sources 

• Control access consistently across your estate 

Without this, AI tools can surface sensitive data to the wrong users. 

How should we handle edge cases like BYOD or isolated systems? 

Two scenarios came up a lot as questions: external users and isolated applications. 

For BYOD users (contractors, volunteers): 

• Follow a Zero Trust approach 

• Enforce MFA and device registration 

• Use Intune app protection to prevent data leakage onto local devices   

For isolated systems (like HR platforms): 

• Keep them deliberately separate if required 

• Integrate via controlled, one-way pipelines 

• Use anonymised or filtered data when connecting to OneLake 

What roles do DBAs and data teams play in Fabric? 

Fabric doesn’t remove responsibilities; it shifts them. 

• DBAs move towards capacity planning, cost management, and performance optimisation 

• Data teams can work more flexibly, using tools like notebooks for Python and R alongside Power BI  

It’s a broader role, but arguably a more strategic one. 

After the session, Rana shared her perspective:

Security in Microsoft Fabric has to be treated as a core design principle rather than something applied later. The general availability of OneLake security at the beginning of May marks a significant milestone, as it introduces a more consistent and unified way to enforce access across users, items, and data paths, with capabilities continuing to roll out. As organisations scale their data platforms, this becomes the foundation for maintaining governance and trust. It also has a direct impact on how AI interacts with data, ensuring that insights are generated within the right security boundaries and only surfaced to the appropriate users.

Rana Kamel, Cloud Solution Architect at Microsoft

Security in Fabric isn’t something you layer on later. It’s something you design from day one. Get that right, and everything else becomes easier, from scaling your data platform to introducing AI safely and with confidence. 

Missed the webinar?

If there’s information we’ve not covered here, or you simply want to learn even more about security and governance in Microsoft Fabric, you can access the on-demand recording to watch the webinar back in full.

Access the webinar

Written by Dan Knott, Data and AI Practice Lead 

Technology alone won’t transform your organisation. If it did, every business with a reporting tool, a data warehouse or an AI pilot would already be an industry leader. 

You can’t buy a data culture. And that’s exactly where most organisations go wrong. 

They expect that a platform, like Microsoft Fabric (powerful as it is), will magically create alignment, consistency, and better decision making. But real cultural change doesn’t come in a box. It comes from people: how they work, how they think, and how they use data to drive the business forward. 

In this follow-up to our Blueprint of a Great Data Culture, I’ll delve into the practical steps IT leaders can take to embed data into the fabric of their organisation and highlight the common pitfalls that stall so many initiatives. 

The 3 Key Mistakes That Hold Businesses Back 

Many organisations that think they’re becoming data driven, but in reality, never quite get there. There are a few common pitfalls that I’ve seen time and time again.  

1. Treating data culture as a technical project: Data culture isn’t a reporting rollout, a BI project, or an AI pilot. Those are outputs. Culture should be embedded in the way of working.  
2. Missing business involvement from the start: Building a data culture will not be possible unless everyone in the business is bought in. Without cross department input, teams are not invested and it can feel like another unnecessary process. 
3. ‘This is how we’ve always done it’ mentality: Gutfeel decisions. Spreadsheets saved on desktops. Siloed versions of the truth. Culture change means letting go of old habits, and that can’t happen without intentional support. 

Practical Steps to implement a strong data culture  

Creating a data culture begins long before platforms and dashboards. IT leaders must shape and secure a business-wide mandate for change. Here’s how: 

Step 1: Start with Leadership and Vision 

Senior leaders must be the ones driving and backing the cultural shift. Leadership support empowers IT teams by enabling them to work seamlessly without resistance. They also encourage the company-wide adoption that is critical for success. 

Key actions to take: 

Build a shared vision with the C-suite 

• Run a workshop with key business leaders to help align data initiatives with business goals. Consider cost reduction, customer growth, compliance and operational efficiency. 

Translate the vision into a business-backed roadmap 

• Build a structured roadmap that demonstrates key milestones, from quick wins to long-term business outcomes. Each milestone should be measured and tracked to a KPI to ensure successful review.  

Secure executive sponsorship 

• Dedicated exec sponsor(s) should champion the change publicly in various forums to reinforce expectations as well as model the intended behaviours you want employees to adopt. They should also ensure that there are adequate resources allocated to the initiative.  

Step 2: Establish a Single Source of Truth 

Every organisation struggles with spreadsheet chaos. It’s easy for different teams to manipulate the same numbers in different ways, potentially coming to wrong conclusions. Most critically, having a trustworthy data platform lays the groundwork for AI readiness. If the data isn’t accurate and trustworthy, neither will the AI be. It’s that simple. 

Key actions to take:  

Identify and prioritise core data domains 

• Start with areas where inconsistent data causes the most friction. This could include customer master data, finance/forecasting, marketing attribution and service delivery metrics. Then run a data audit to understand where the data lives, who owns it, how many versions exist, and how it’s currently used (and misused). 

Design a governed, accessible architecture 

• This typically includes implementing a unified data platform such as Microsoft Fabric, defining ownership and ingestion processes, as well as controlling access where necessary. There should also be a built-in data quality and validation processes. But governance doesn’t have to be heavy-handed. You need to start with core principles, not long policy documents. 

Clean the data before surfacing it 

Before exposing dashboards, you should… 

• Fix critical data quality issues 

• Validate key metrics with business users 

• Document known limitations 

• Test data accuracy with “friendly sceptics” in the business 

Step 3: Build Data Literacy Across the Business 

A sophisticated data platform is meaningless if people don’t know how to use it, or don’t share the same language. Data literacy isn’t about teaching everyone Python or SQL. It’s about ensuring people can understand the metrics, interpret the data, and apply insights in their role. 

Key actions to take:  

Create a business glossary 

Start with essential, high-impact terms such as:  

• What does “Gross Margin” mean here? 

• When does an Opportunity become Qualified? 

Co-create definitions with each business function. Host them in a central, easy-to-access place. 

Run role specific training 

Avoid generic “Power BI training for everyone,” And instead design learning journeys by role: 

• Exec team: reading dashboards, challenging assumptions 

• Sales: understanding pipeline metrics 

• Marketing: attribution logic, campaign performance 

• Finance: forecasting and scenario modelling 

• IT: governance, access, troubleshooting, lineage 

You should also make data literacy part of your company’s induction process. This makes data driven behaviour a default, not a bonus. 

Encourage “data ambassadors” 

Identify individuals who naturally ask good questions and understand the tools. These are people who can influence their peers and support with questions. 

Step 4: Embed Data into Everyday Behaviour 

A culture only forms when new behaviours become habit. It needs to be embedded in employees’ regular ways of working to be effective.  

Key actions to take:

Integrate data into existing rituals 

Rather than creating new processes: 

• Use dashboards for monthly performance reviews 

• Review KPIs at weekly team meetings 

• Incorporate data into quarterly planning 

• Make reports the default starting point for decision making 

Standardise on agreed tools and dashboards 

• Nothing kills culture like fragmentation. Ensure your entire organisation is aligned by making it clear which dashboards are “the single source of truth” and decommission redundant spreadsheets and shadow systems. 

Use Power BI and Fabric to democratise access 

• Give teams self-serve analytics where appropriate, but with standardised definitions, pre-built data models, and simple, clear visualisations. Self-serve only works when the foundation is governed. 

Step 5: Start Small, Celebrate Wins, and Iterate 

Cultures shift gradually. Start small, and celebrate where it is going well. 

Choose small, visible use cases first 

Good examples: 

• Reducing forecast discrepancies between Sales and Finance 

• Improving campaign reporting accuracy in Marketing 

• Reducing manual spreadsheet reconciliation in Operations 

Quick wins build belief. 

Create a feedback loop 

You need that feedback loop… a great data culture never stands still. 

Implement: 

• Monthly data council meetings 

• Dashboards tracking data quality 

• Suggestion channels 

• Regular retrospectives after report launches 

Publicly celebrate success stories 

People adopt what they see being rewarded. You should showcase wins including: Who used data to solve a problem, Where a decision changed because of insight and what the key outcome was.

Iterate continuously 

Your data culture is a living system, not a project: 

• Update definitions as the business evolves 

• Add new use cases 

• Improve data models 

• Retire reports that no longer serve a purpose 

• Continually refine governance 

Culture compounds through iteration. 

Ready to strengthen your data culture? 

If you’re looking to build a data culture that drives decisions, unlocks AI readiness, and aligns your entire organisation, we can help you. Cloud Direct works with leaders who want to break through old ways of working. To build something stronger, smarter, and more sustainable.  Reach out to us using the form below.

If you’re searching for more information to help plan your AI journey, our data and AI playbook can help to guide you.

In ‘The Data & AI Readiness Playbook’ we explain how you can unlock the value of your business data. Quality data is crucial to the effectiveness of AI. Here we look in more detail at the importance of data culture and the practical steps you can take to build a great data culture.    

Data is perhaps the single greatest asset of the modern business and the bedrock of successful AI initiatives. But the quality of that data is critical to success.  

The ROI of investing in data quality

Already convinced of the importance of quality data? Then skip ahead to ‘Developing a great data culture’, otherwise read on.  

When we refer to quality data, we mean that it is accurate, complete, consistent, timely, valid, unique, and reliable. Armed with that, both human and artificial intelligence can arrive at the decisions that deliver real value.

Matthew Ebo is Assistant Strategic Insights Manager at Lloyds Banking Group and a contributor to ‘The Data & AI Readiness Playbook’. Matthew explains, “The cost of AI is an investment, but one that pays off in saved time, better decision-making, and automation of repetitive tasks”.  

Developing a great data culture is an investment. An investment in the raw materials for transforming how an organisation operates, competes, and grows. For the business, this can drive value in many ways: 

• Better, evidence-based decisions enabling more effective strategies, faster responses to market changes, and improved stakeholder buy-in for new initiatives. 

• Increased innovation and agility as the organisation becomes better at identifying new opportunities, anticipating trends, and proactively adapting to changes.  

• Improved operational efficiencies as data insights help identify inefficiencies and optimise processes and resource allocation.  

• Greater competitive advantage gained from better anticipation of customer needs, reaction to industry shifts, and speedier innovation.  

• Enhanced customer experience through superior analysis of customer data, to tailor products, services, and communications to better meet customer needs.  

• Boosted employee engagement as staff become more engaged, and invested in the success of the business, by using data in their roles. 

• Trustworthiness – data driven cultures tend to be open about business decisions improving trust amongst employees.

A great data culture enables smarter decisions, drives innovation, enhances customer and employee experiences, and provides a competitive edge. 

So how do you get one?  

Developing a great data culture

We know that AI initiatives are only as good as the data they access, so data quality is key – and great data follows a great data culture. 

Data quality isn’t solely the responsibility of IT: it’s everyone’s responsibility. But it is IT’s responsibility to encourage a culture in which everyone cares about data quality“, explains Dan Knott, Data & AI Practice Lead, Cloud Direct.

A great data culture is one where data is valued by an organisation and its people, becoming a key part of daily operations. Here are eight key aspects of a great data culture.  

1. Leadership commitment and example
As with most aspects of culture, data culture needs to be led from the very top. Senior management need to be actively involved in communicating the importance of data to organisational success. This needs to be supported with investment in the resources, training, and technology to support data initiatives.

2. Data-driven decision making
It needs to become the norm that decisions, at all levels, are based on data and analysis and not just intuition or hierarchy. Leadership and example are an important part of this. It also means establishing processes that make data analysis a standard part of planning, operational, and problem-solving activities.

3. Data access and democratisation
For IT this means ensuring that employees have easy, secure access to data and analytics tools. Employees should be able to freely share data internally, within appropriate privacy and security parameters, to enable collaboration.

4. Data literacy and training
Key to all this is an ongoing investment in employee’s data awareness and skills, ideally with training tailored to different roles and data personas. This should equip staff with the skills to interpret analytics, know when to use data, and to ask the right questions about data quality.

We’re looking at AI in the same way as other tools which become part of the workforce’s toolbelt, so we have to provide the same level of training for it. It’s not only imperative that the right skills are gained, but that they are regained as time goes on“, explains Mike Downing, Chief Technology Officer at insurance nonprofit WPA

5. Continuous learning and improvement
Alongside formal training initiatives, organisation’s need to develop a mindset of ongoing learning, experimentation, and adaptation – constantly utilising data to refine strategies and processes.

6. Trust in data quality
Underpinning this, there needs to be organisation-wide confidence in data accuracy and consistency. People know where data comes from and how reliable it is. This requires robust data governance, clear data lineage, and transparent processes for data validation and error correction.

7. Collaboration and communication
Success will be evidenced through open communication and collaboration around data, with teams working together to solve problems and share insights. This should be openly encouraged. It is important that this is also supported by establishing a common ‘data language’, so everyone communicates effectively.

8. Accountability and measurement
The final aspect of great data culture is one of transparency and it is an extension of point two. Clear goals and metrics are set for data initiatives, with progress tracked and results linked to business outcomes. With performance KPIs relating to data usage built into everyone’s evaluations.

Together, these elements can help you to build a great data culture that will deliver demonstrable value to your organisation and support effective use of AI.  

Next Steps

Download a copy of The Data & AI Readiness Playbook to learn more. You’ll discover how others are preparing for and using AI, and a seven-step process to unlock the value of your business data.  

It’s becoming increasingly apparent that artificial intelligence will be integral to how organisations operate effectively and remain competitive. But responsibility is a topic that regularly rears its head, and the question of how you use AI responsibly isn’t one purely for IT but for your organisation’s executive leadership. Here we consider how to benefit from AI, while remaining true to your organisation’s values, obligations and stakeholders.  

Much is written and spoken of AI’s power to drive business transformation, efficiency and innovation. But as the saying goes, ‘with great power comes great responsibility’.

Using AI responsibly isn’t just about regulatory compliance. It’s about trust, safeguarding reputation, and ensuring that AI strengthens rather than undermines the organisation’s values and purpose. There are three important topics to consider – People, Planet, and Policy.

People

Let’s start by confronting the really big question: jobs. We’ve all seen and heard carefully worded references to AI’s labour-saving capabilities. It does less work, it means fewer workers, but does that also mean redundancies? This question needs considering, carefully, at a very senior level, and very early on.

They’ll need to know the expected time savings and whether these affect fractions of roles or entire roles. You should also consider timeframes in each area, and how these compare to natural attrition, retirement and contract expiration timescales. They’ll also need to know recruitment pipelines, so hiring can be slowed or redirected rather than abruptly frozen, as well as redeployment opportunities and the skills required for new or expanded roles such as AI oversight, data literacy, and creative problem-solving.

This will impact much of what follows.      

Enablement, not displacement

Responsible AI should augment human judgement and not replace it – freeing people from repetitive work. But to achieve this, it needs to be accompanied byreskilling and digital literacy programmes that enable employees to work effectively with AI systems. Success should be measured in terms of human productivity and satisfaction, and not headcount reduction.

Transparent and ethical

Everyone involved with AI, from developers to decision-makers, must understand what AI can and cannot do. Build a culture of AI literacy and ethical awareness supported by specific training on responsible data use, bias awareness, and explainability. Employees using AI outputs should be able to interpret and justify its decisions, especially in regulated sectors. Staff must appreciate that humans remain accountable for AI-assisted outcomes and feel confident challenging algorithmic decisions without recrimination.

Inclusion and fairness

Similarly, fairness and inclusion must be embedded in your use of AI. These systems will typically maintain or increase any biases in training data, so utilise diverse teams in AI design and validation. Train models with diverse data sets and monitor for bias, especially in HR, credit, or customer-facing use cases.Treat governance of AI fairness with the importance of a workplace equality and diversity issue, rather than a technical issue.

Planet

AI’s benefits should also be considered in the context of its environmental impact and sustainability. During training AI models can consume significant energy, and operationally AI infrastructure has a significant carbon footprint. But with the right actions, this can be mitigated.

Opt for energy-efficient architectures

Data centres powered by renewable energy, with liquid cooling, and using energy-optimised GPUs (Graphic Processing Units) and ASICs (Application-Specific Integrated Circuits) are more energy efficient. Also consider scheduling AI workloads to optimise power use.

Actively manage your technology lifecycle

Using cloud and hybrid models can allow you to dynamically scale, without having an over-provisioned on-premises infrastructure. Apply sustainability principles to AI hardware: responsibly sourcing, refurbishing and/or reusing, and recycling at end-of-life.

Use AI for sustainability

‘Planet’ doesn’t just mean mitigating AI’s environment impact. AI can also make a positive contribution towards meeting corporate sustainability goals through data-driven energy optimisation, intelligent logistics routing that lowers emissions, predictive maintenance to reduce waste, and carbon accounting.

Policy

A responsible use of AI also depends on robust governance that ensures transparency, accountability, and compliance. A key consideration for the board is who will be accountable for AI ethics and compliance, and how governance can be shown to be effective?

A best practice approach combines collective ownership with clear executive accountability. It is likely to blend existing structures with some new, specialised capabilities. This might take the form of a Chief Information Officer or Chief Digital/Technology Officer with primary accountability, working with a cross-functional AI Governance Board. This would include Technology, Data, HR/People, Legal, Compliance, Risk, Operations, your ESG (Environmental, Social, and Governance) team, and business unit leaders.

This will provide the basis for effectively actioning the following. 

Establish an AI governance framework

Determine the principles which will guide your use of AI. These need to be consistent with your organisation’s values and risk appetite and will typically encompass fairness, transparency, accountability, privacy, and sustainability. Bear in mind that different contexts may require different ethical considerations – what’s appropriate in one area may not be in another. AI ethics will touch IT, legal, HR and compliance so ensure that there is clear ownership within and across these areas.

Control and oversight

Integrate AI risk management into existing risk frameworks, with a focus on model validation, auditability, explainability, and version control. Track who built which model, with what data, and how it is performing. Require human-in-the-loop oversight for all critical decision and systems.

Regulatory alignment

There will be external interest in your AI use from regulators, customers, investors and other stakeholders, so aim to stay ahead of expectation. There is an EU AI Act, with most provisions applying from August 2026, and a UK AI Assurance Framework. The Information Commissioner’s Office has provided AI guidance, with sector-specific guidance expected in several areas (like from the FCA in financial services). Maintain audit trails for AI models, data lineage, and decision logic to satisfy auditors and regulators.

But, above all, be transparent about how AI is used, governed, and improved.

A final thought

Using AI responsibly requires deliberate, pre-emptive leadership. It means ensuring that AI use aligns with organisational purpose, is trusted by employees and other stakeholders, and contributes to sustainable growth. Many will do this badly, but those that do it well can successfully position their organisations as trustworthy and responsible innovators.

Cloud Direct can help you successfully benefit from AI in a real and responsible way. Request a call with a subject matter experts through the form below.

Used correctly, AI can transform your organisation’s use of data – and The Data & AI Readiness Playbook gives you a seven-stage process for doing this successfully. However, user adoption can make the difference between success and failure for an otherwise well-thought through initiative. Here we look at the issues you might face and how you can overcome them.

As technologists, we want to believe that it’s our choices and actions that make the real difference. But it’s not tools that transform business, it’s people. Survey after survey shows that it’s user adoption that is the real key to the success of your transformation project“, explains Dan Knott, Cloud Direct Data & AI Practice Lead. 

Motivating employees to use new tools effectively, and measuring and managing this, is the difference between costly failure and transformative success. But before we look at how you do this, let’s consider why this might be necessary.  

Why bother driving adoption?

As humans, we are essentially creatures of habit. Neurologically, our brains are wired to form habits, and they account for roughly 40-45% of our daily actions.  

A new process or technology often requires us to change our behaviour – and it may seem easier to carry on as before. So, there’s also a level of inherent resistance to change that needs to be overcome.   

However, as humans we also have a remarkable capacity for learning and adaptation, but we need to see the benefit of change. Subconsciously we’re asking ourselves ‘what’s in it for me?’     

Where AI is involved there’s often an added mix of practical, emotional, and ethical concerns that also need resolving. Will it take my job? Will AI make decisions for me? Can I trust it? Will I be held accountable for AI errors? Will it be used to monitor me?  

Unless and until people know otherwise, these are valid concerns over what AI means for employees’ roles, job security, and daily experience. 

Tackling AI anxiety and engaging employees in change requires much more than good comms. It requires empathy, clear communication, visible leadership, and a people-first approach.  

Initiatives with strong adoption and change management processes are six times more likely to succeed.

According to a survey by change management specialists Prosi.

How to maximise adoption

If we’re being honest, AI is a big shift and for many employees it will be a big deal. So, let’s treat it like one. Here are seven key elements of driving a successful adoption. 

1. Build Change Management in from the start 
   • Establish a formal change management programme from day one. Change Champions will play a vital role as advocates for changes and as providers of peer-to-peer support. Alongside this you need a communication strategy that clearly explains ‘why’ change is occurring as well as the ‘what’ and ‘how’.  
2. Acknowledge and address user concerns  
   • Openly acknowledge anxiety and recognise that resistance is natural and expected. Different groups of employees will have differing needs, and you may need to create safe spaces for concerns to be voiced. Address common objections head-on with clear explanations and evidence. Often resistance stems from job fears, so be transparent about how roles will evolve rather than disappear. 
3. Explain the benefits  
   • Users like clear, tangible benefits. While it’s important to explain how AI is important to the business, answering ‘What’s in it for me?’ is more important in driving adoption. Position AI as an enabler (rather than a threat) and emphasise how AI can remove tedious tasks and support better decisions. Help people to see how it augments human capability, rather than replacing it. Ideally, tailor the messaging and provide use cases that are relatable to different types of roles.  
4. Involve employees early and often 
   • Inclusion turns resistance into ownership, so invite input into how AI can be used in specific departments. Directly ask: “How and where could AI help you most?” Regularly communicate progress and share real examples of how employees are using AI to make work easier or more rewarding. Since personal stories build trust far faster than tech demos, personalise examples to celebrate people, and not just the tech. For example, ‘How AI is helping Julie to produce monthly reports in half the time’. 
5. Provide practical, role-relevant training 
   • Focus on hands-on, scenario-based training, rather than theory. Deliver the training in as many different bite-sized formats as possible to suit different learning styles: lunch-and-learns, online tutorials, quick reference guides, drop-in sessions. As a rule, when people can see the benefit, they want to know more.  
6. Lead by example 
   • If leaders are visibly seen to be using AI, others quickly tend to follow. Not only does leaders’ use signal AI’s importance, but many of us learn working behaviours and practices from those above us. Nothing builds adoption like seeing your boss on the tools. 
7. Measure and iterate 
   • Success in user adoption requires it to be seen as an ongoing journey rather than a one-time event. Track quantitatively (usage rates, task completion times, errors) and qualitatively (user satisfaction and feedback) and use this to identify adoption barriers and continuously improve. 

Although this may seem like a lot of extra work, adoption is likely to be the single biggest determinant of project success. A good partner will help you to engage and involve internal comms, HR, change management and other colleagues that will help you achieve all this.  

Next steps

Download a copy of The Data & AI Readiness Playbook to learn more. You’ll discover how others are preparing for and using AI, and a seven-step process to unlock the value of your business data.  

Data quality is fundamental, with the cost of bad data running into millions for UK businesses. Here, we look at how and why poor data is costing your organisation dearly, and the practical steps you can take to improve data quality. 

Bad data is costly. Research by analysts Gartner puts the average cost to organisations at a staggering $12.9 million a year – around £10 million. Whether that figure seems high or low to you will depend on a variety of organisation-specific factors, but the key point is that poor data is costly.  

As we become more AI-reliant, that cost is likely to increase. And for most, it’s completely hidden.   

Six ways data deficiencies could be costing your organisation

Poor data typically impacts organisations through operational waste, revenue leakage, and strategic shortcomings.   

1. Lost revenue  
   Data inaccuracy can result in poor decisions, causing lost sales, poor targeting, and underachieving campaigns. Old or incomplete customer data can mean missed upsell or cross-sell opportunities.  
2. Wasted time  
   Employees waste hours finding, verifying, or correcting information.  
3. Increased risk  
   Data errors cause GDPR and regulatory breaches, possibly incurring fines and certainly requiring time-consuming corrective action. Risk models that use poor data will fail to flag preventable issues. 
4. Negative customer experience 
   Incorrect contact details, preferences, or histories lead to irrelevant messaging or service errors, which undermine customer confidence, damage brand reputation, and add to customer churn.  
5. Operational inefficiencies 
   Errors in billing, shipping, or inventory management cause rework, returns, and delays – all of which carry a cost.  
6. Strategic failings 
   Decisions based on flawed insights can misallocate investment, degrade valid opportunities, ignore key risks, and cause growth strategies to fail. 

Why data matters more than ever 

As John Doublard, CTO at Oak Group and one of the industry contributors to ‘The Data & AI Readiness Playbook’, notes: “The future is data-driven.” 

“AI can only drive genuine business value when it addresses real business issues AND is fuelled by valid data,” explains Cloud Direct’s Data & AI Practice Lead Dan Knott. 

Understanding the limitations of your data, which parts need improving, and how to make those improvements is key to this.  

Eight practical steps for data quality improvement

Improving data quality isn’t just the responsibility of IT – although there are technical actions that will help – but something for the whole organisation.   

‘When you’re about to invest heavily in becoming a data-driven business, you need to make sure that the data you’re working with is going to provide for you and not cost you more’.

1. Define what ‘good’ means 
   If you’re working with bad data, you’ll get bad outcomes, so establish clear data quality dimensions. The Data & AI Readiness Playbook outlines the importance of accuracy, completeness, consistency, validity, uniqueness and timeliness. Add to this any business-specific requirements, such as capturing an accurate address or company registration number for credit checking.  
2. Perform a data audit 
   Identify your key data assets, such as CRM/ERP and finance data, and assess current data quality. Prioritise high-impact areas and make use of tools to help you highlight errors, duplicates, gaps, or inconsistencies.
3. Fix existing issues  
   Involve data owners (see below) with data cleansing. The objective is to de-duplicate records, fill in missing fields, standardise formats, and identify old and obsolete data for archiving and removal. Tools like Microsoft Purview and Power Query can helpfully automate some of the work. 
4. Embed good data governance practices 
   Appoint data owners and data stewards who will be responsible for data quality in key areas. Involve them in creating policies for data entry, usage, retention, and updates, and make use of Microsoft Purview, or similar, to enforce policies and manage access, lineage, and classification.
5. Educate and engage users 
   Data quality needs to be seen as a shared responsibility, and not IT’s problem alone. Run awareness campaigns or training to help users understand their role in maintaining good data. Within this, celebrate success by showing how clean, reliable data is enabling better decisions and results, and continue running campaigns until data quality is embedded in the organisation’s culture. 
6. Improve data entry at the source 
   You don’t want to be continually drawn into corrective action, so employ measures to improve data entry at the source. Add validation rules, drop-downs, and formatting controls where data is input, and ensure that departmental training covers the importance of entering accurate data. Wherever possible, automate data capture utilising integrations, forms with logic, and barcode scanners. Improving data quality requires both technical and cultural change. Start with what matters most, fix the root causes, and embed quality into everyday workflows. Over time, better data leads to faster decisions, happier customers, and lower costs.

Next Steps

Download a copy of The Data & AI Readiness Playbook to learn more. You’ll discover how others are preparing for and using AI, and a seven-step process to unlock the value of your business data.

Written by Cloud Direct CTO Paul Sells

In today’s fast-paced digital economy, artificial intelligence (AI) is a business imperative. From automating routine tasks to uncovering deep insights from data, AI is transforming how business operate, compete, and grow. 

But here’s the challenge: while many companies are eager to embrace AI, few have a clear strategy for doing so effectively and sustainably. That’s where an AI Centre of Excellence (AI CoE) comes in. 

What is an AI Centre of Excellence? 

An AI CoE is a dedicated team or function within a business that leads and governs AI initiatives across departments. It’s an internal AI consultancy, with the aim of bringing together the right people, processes, and technologies to ensure your AI investments deliver real business value. 

Rather than scattering AI efforts across siloed teams, an AI CoE provides a centralised hub of expertise, best practices, and reusable assets. It helps your organisation move from isolated experiments to enterprise-wide impact. 

Why Do You Need an AI CoE? 

Implementing AI isn’t just about buying tools or hiring data scientists. It’s about embedding intelligence into the fabric of your business. Without a structured approach, AI projects often suffer from: 

• Lack of strategic alignment: Teams build models that don’t solve real business problems. 

• Duplication of effort: Different departments reinvent the wheel with similar use cases. 

• Talent bottlenecks: Skilled AI professionals are spread too thin or underutilised. 

• Governance gaps: Ethical, legal, and compliance risks go unmanaged. 

An AI CoE addresses these challenges head-on by providing a unified framework for AI adoption. It ensures that your AI efforts are not only technically sound but also strategically aligned, ethically responsible, and scalable. 

Benefits of an AI CoE 

Establishing an AI CoE isn’t just a technical investment—it’s a strategic one. Here are some of the key benefits: 

1. Faster Time to Value with centralised expertise and reusable assets, AI projects can be delivered more quickly and efficiently. 
2. Improved ROI by focusing on high-impact use cases and avoiding duplication, the CoE ensures that AI investments generate measurable business outcomes. 
3. Stronger Governance the CoE provides a structured approach to managing AI risks, from data privacy to algorithmic bias. 
4. Scalable Innovation as AI maturity grows, the CoE helps scale successful pilots into enterprise-wide solutions. 
5. Empowered Workforce through training and support, the CoE builds a culture of innovation and continuous learning. 

Real-World Example: AI Centre of Excellence in a Law Firm 

Picture a growing law firm under increasing pressure to manage an ever-expanding volume of contracts, case files, and compliance documentation. Different practice areas are starting to explore AI—contract review here, legal research tools there—but it’s all happening in pockets, with little coordination. The result? Duplication, inefficiency, and missed opportunities. 

This is exactly where an AI CoE can make a measurable difference. 

By putting an AI CoE in place, the firm can: 

• Spot the highest-value opportunities, like automating contract intelligence and streamlining legal research and other use cases that benefit the entire firm, not just individual teams. 

• Bring together the right mix of people including legal operations, IT, data scientists, and senior partners, to co-create a firm-wide solution for reviewing, classifying, and extracting key terms from contracts.  

• Scale smarter, by adapting AI models across departments for consistent results and slashing time spent on manual reviews. 

• Build trust through governance, embedding legal ethics, client confidentiality, and compliance into every AI workflow. 

• Invest in training and change management, so that associates, paralegals, and support teams feel confident using these tools—not threatened by them. 

The impact? Engagements move faster. Risk reduces. Compliance becomes easier. And your top legal talent gets to focus on strategic advisory, not repetitive admin. Ultimately, it’s about delivering better outcomes for clients, and creating space for the kind of work that grows the firm. 

When external expertise is brought in to fill internal skill gaps which is often the case with AI projects, the AI CoE acts as the strategic bridge. Allowing alignment between external parties efforts and the firm’s goals, standards, and governance frameworks. 

How to Get Started

Ready to build your AI CoE? Here are the first steps: 

1. Secure Executive Sponsorship 

An AI CoE needs strong backing from senior leadership. Showcase how AI aligns with your strategic goals, whether that’s improving customer experience, reducing costs, or driving innovation. 

2. Define the Scope and Structure 

Decide whether your CoE will be centralised (a single team), federated (embedded in business units), or a hybrid of these. Start small with a few high-impact use cases and scale over time. 

3. Assemble the Right Team 

Look for a mix of technical and business talent. You don’t need a huge team to start—just the right people with a shared vision. 

4. Establish Governance 

Define clear policies for data usage, model validation, and ethical AI. Set up review boards or steering committees to oversee major initiatives. 

5. Invest in Tools and Infrastructure 

Choose platforms that support collaboration, version control, and model deployment. Cloud-based solutions often offer the flexibility and scalability you need. 

6. Measure and Communicate Impact 

Track KPIs such as cost savings, revenue uplift, or customer satisfaction. Share success stories to build momentum and secure ongoing support. 

Final Thoughts 

AI is a journey not just a one-off project. And like any journey, it needs a map, a guide, and a destination. An AI Centre of Excellence provides all three. 

By investing in a CoE, you’re not just adopting AI—you’re building the foundation for a smarter, more agile, and more competitive organisation. 

So, whether you’re just starting out or looking to scale your AI efforts, now is the time to consider: Do we have the right structure in place to make AI work for us? 

It’s no secret that businesses are producing more data than ever, and you will only be able to drive greater efficiencies and enable innovation once you have a clear strategy with the right tools in place.

Earlier this year, Microsoft launched Fabric, their all-in-one data solution, for general availability. Microsoft Fabric combines some of Microsoft’s most powerful tools, such as Data Factory, Synapse Analytics, Data Explorer, and Power BI into a unified, cloud-based platform to help simplify your data workflow. The combination of the tools will enable you to innovate with AI safely and securely by managing your data in a single user-friendly platform.