Every organization needs AI capability, but not everything needs to be built from scratch.

You need to harness your existing people and understand their capabilities. Can your existing data analytics and engineering capabilities be upskilled in modern techniques and deployed to create value and differentiate your business? The skills gap is rarely where leaders expect it. What's underestimated: data engineering, integration engineering, product management, and change management.

Make the build, buy, or partner decision workflow by workflow, based on its ability to create value and differentiate your business. Buy talent that is common in your industry and is easier to recruit for. Build your top talent’s capabilities in areas that can differentiate your business. Partner when you need differentiation faster or cheaper than you can build or hire the talent.

Where possible, hold buy decisions on technology until your people have built and implemented the first few use cases and proved the long-term technology choices. With this knowledge, your employees can intelligently assess vendor claims and negotiate more effectively.

Regardless of where the capabilities sit, the key is to point them to the value drivers we discussed in Strategy. Embed this in your business by pairing technical and business people on the same teams, so AI expertise sits where value is created rather than in a separate function. An AI team that never visits the plant floor will build models that miss what matters most.

How you structure AI capability determines what actually gets delivered and how well initiatives can scale. Most organizations need a small central team. Whether you call it a transformation office, a PMO, an accelerator, or something else, its job is the same: help executives choose the right priorities, set those priorities up to succeed, share what is learned so each business unit does not have to learn it alone, and hold the business accountable for cost and ROI [link to Governance section pending].

But central coordination should not equate to central ownership. Once a priority is chosen, it belongs to the business leader accountable for that part of the business. If it is a supply chain initiative, the supply chain director owns it, because that is the person who is best placed to land the change in their own area successfully.

Be clear about who runs what and who is accountable for the outcome. Reporting lines carry a powerful signal. A technology leader reporting to the CFO will likely be asked to manage costs, while one reporting to the CEO can be positioned to drive growth. In large-scale manufacturing, this has long been understood: at General Motors, the CIO reports directly to the CEO, reflecting a view that technology is central to competitive strategy.

Start well before deployment with an honest read on readiness: leadership alignment, cultural receptivity, capability gaps, and process dependencies. The earlier the resistance surfaces, the more opportunity there is to address it. Employees also need a clear story about why this is happening, what it means for their role, and what that role looks like in twelve months. People can handle change. What they struggle with is uncertainty.

Whether they get on board comes down to three things: autonomy, control, and belonging. Get those right, and adoption builds its own momentum. Get them wrong, and no amount of training will compensate.

Measure adoption as rigorously as you measure technical performance. Time saved only matters when the capacity is genuinely redeployed, so track that rather than the hours a model theoretically freed up. Redesign incentives at the same time, since people measured on manual activity have no reason to hand that work to an agent.

Ultimately, change should be a permanent capability rather than a one-time need, because the next pivot is already coming. Businesses that have agility when change is most urgent or lucrative are the ones that built the muscle before they needed it.

The CEO of a technology services business needed to show the market a credible path to a major EBITDA improvement after carving out a significant part of the company. The company’s own marketing had even promised that the business was already AI-led.

It wasn't. There was an AI team and a great deal of data, but both sat apart from the business, disconnected from where value was created. All resources were there, but sat effectively untapped.

The immediate problem was structural. Reshaping an organization of more than 10,000 roles meant taking out cost without cutting into the capabilities the remaining business depended on. Ownership across the business was fragmented, decisions were made by consensus, and no leader had clear authority to reshape their own function.

We designed a change management process that would integrate leadership's new vision with the reality of day-to-day work across the business. From the top, we took the CEO’s EBITDA target and mapped it to savings hypotheses for each function, benchmarked against comparable businesses.

We also mapped the work task by task across each function, surfacing improvements within teams and across them. Every initiative went into a single tracker with a value, a cost to achieve, a timeline, and a named owner, each tested by Finance before any savings counted.

Then we sat down with each manager to agree how roles needed to change to achieve the business’s objectives. Working with HR, we sequenced the changes and the communications around them, including the notice and approval steps local employment law required, so people knew what was changing and when.

Along the way, we also built what the marketing had promised long before we arrived. The business’s commercial data had been largely untapped, so we added a commercial workstream. Working with the company's existing data scientists and its commercial leaders, we built a machine learning model that predicted which deals in the pipeline the company was most likely to win, with 82% accuracy. We set it up to run in the company's own systems, operated by its own data scientists, so the capability stayed with the business rather than leaving with us. We then designed two new teams to put the insights to work: a Revenue Win Room to prioritize pipeline based on the new model and a Deal Desk to execute pricing and terms on those deals.

The CEO now had a credible path to show the market, with $230 million in annualized run-rate EBITDA improvement in place or underway and roughly $60 million already realized. The AI capability the company’s marketing had promised was finally pointed at revenue, with the structures designed to put it to work. The company later brought us back to extend it, from which deals to win to which customers to sell more to.