Q&A: Where are you on your AI journey?

• Start with value, not adoption: The difference between businesses that get value from AI and those that do not is rarely the technology. It is whether work, roles and accountability have been redesigned around it.
  
  
• Build the right foundations: High-quality data, clear governance and fit-for-purpose guardrails determine how far and how safely AI can scale.
  
  
• Make AI observable: As AI becomes embedded in core processes, businesses need to be able to show what AI did, what people did and how each outcome was produced, particularly in regulated or high-trust industries.

Many organisations have moved beyond asking whether they should use AI. The harder question is why so few convert early experimentation into measurable business value. In most cases, technology is not the barrier.

Datacom Consulting Director Ash Valanidas says most organisations do not struggle with using AI as a tool. They struggle to convert that capability into meaningful outcomes. Instead of redesigning work around AI, many simply layer AI onto existing processes, limiting the value they achieve.

In this Q&A, he discusses how businesses can assess their AI maturity, move from education and experimentation into operating, integrating and scaling AI, measure the value of Microsoft 365 Copilot and judge whether they are ready for the next stage.

AI maturity is rarely uniform across an organisation.

I think that most organisations will sit across several stages at once. With AI journeys being a part of business operations, firms will be educating their people, experimenting with tools, operating proof-of-concepts, integrating AI into core operations and eventually, scaling adoption across the organisation. It’s quite reasonable to be advanced in one function and still experimenting in another.

I also see two emerging corporate personas in AI value. The first is those still at the education stage.  They are interested in frontier AI but need further education on where it could create value and how they should approach it. I think this is often the case in highly regulated, publicly accountable or resource-constrained organisations, such as parts of the public sector, financial services, health and NGOs, where investment must clear higher thresholds of trust, evidence and affordability before it scales.

The second is those that are experimenting, but have become stuck before they can operate, integrate or scale. AI is clearly in demand in their engine room and on the customer front line. They have spent a year or more running pilots, trained employees and handed agentic tools to drive personal productivity, yet they struggle to move beyond experimentation.

If I am speaking to a company that sees themselves in this persona, I start to dig deeper on their governance, compliance, culture and cost, and very quickly it becomes apparent that what’s missing isn’t the technology, it’s an operating model to support the AI ambition.

One of the biggest mistakes is measuring usage instead of impact.

I’ll be very clear with what isn’t the value metric - the number of licenses bought, logins recorded or prompts entered. Simplistically, value measurement on Copilot, or other tools, will always come back to what business outcomes you are trying to achieve, and whether Copilot is helping you achieve them at a lower cost than the alternative.

I recommend that organisations start with a clear strategy for Copilot, define specific goals, and then set measurable outcomes. For example, a large enterprise rolling out Copilot to 1000 employees must set a benchmark for the time people save each week through better research and faster project work, carry that productivity into total utilisation and then have a clear plan for what to do with it.

As an example, my team at Datacom has been working with the Frontier Copilot stack for a few months, and we’ve been able to radically shift our workflows in consulting cases and market engagement. Over this period, we've built secure agentic workflows that compress work which used to take days of research and QA into a matter of hours. That's freed the team up to put more into R&D and to show customers what's coming next.

Using AI as a productivity tool focuses on helping individuals work faster.

I speak to a lot of people who believe that the height of AI is simple use cases like creating documents more quickly, conducting research more efficiently, and automating routine tasks. Those benefits are valuable, but they tend to remain at the individual level.

True AI-led transformation happens when a business takes a broader view and moves from experimenting to operating and standing up proof-of-concepts that actually work inside the organisation. From there, you integrate the ones worth keeping into core operations and start reimagining the process end-to-end. That means looking across a whole value chain, finding where AI can join up the decisions and hand-offs, and working out where a mix of agents and people in the loop can do more with less.

A university, for example, can use AI to connect the entire student experience, from an initial admissions enquiry, to enrolment, to graduation, to alumni. With that connected, autonomous workflow, a university can make a prospective student an instant offer with a customised enrolment portal and learning experience.

And I think this university story is the goal. It is to move from seeing AI as personal productivity that is scaled across the back office, to a reality where the AI orchestrates an entire value chain.

I would like to urge any reader to take a pause and question how much of their current spend is pouring into speeding up a single task a knowledge worker already handles well and challenge them to think about whether their team is truly taking advantage of AI for transformational productivity or just scratching the surface.

Frontier capability does not create an operating model, but it can provide the foundation one requires.

The Frontier tools are the biggest step change I've seen in solving the trust and resilience problem. By design, they keep a company's AI workflows inside its own perimeter, so you keep tight control over your data, and you can actually see what the agents are doing.

And reflecting on the last 12 months, these features have been on many company wishlists but most tech conferences were stacked with great companies providing a tool for one or two of these use cases, but the integration and completeness remain barriers.

With E7 and the Frontier services, there is a lot of out of the box capability from Copilot, Purview and A365 which will give companies a lot more confidence to experiment and scale as the regulatory context around AI becomes more sophisticated.

The first is attempting to scale AI without a clear strategy or executive sponsorship.

Of the many mistakes I see, a few occur far more often than others:

• Teams implementing AI independently, which creates unnecessary complexity, higher costs and governance gaps, and inevitably forces a pause to define an operating model.
  
  
• Assuming adoption can be driven entirely from the top down, creating an all-too-familiar tension of poor change and operations because the real users were never consulted and the true opportunities probably missed.
  
  
• Failing to govern AI consumption, which will lead to rapidly growing costs, stalled adoption rates and possibly cancelled projects and tools.

I'd add that these three are becoming all too common, so if you're starting out, build your strategic, adoption and cost constraints into the design from day one. I'd also say this for anyone worried they've left it too late... there's a real advantage in being a fast-follower right now. A lot of the hard lessons have already been paid for, and by 2027 the tools will be at scale to buy and configure, and the operating excellence playbooks will have mostly been written for adoption in your AI change programme.

Design with strong foundations and guardrails from the start.

The businesses that make real progress over the next 12 months will be the ones that can explain, measure, and govern the value AI creates. Without that, a business has adopted AI technology, but it has not yet built an AI capability. 

If any of this resonates, or you are weighing up where your organisation sits on this journey, I would genuinely welcome the conversation. Please feel free to reach out to me directly at Datacom. I am always happy to share what we are learning and to help you find the next step that makes sense for you.