Most AI engagements are scoped around an outcome, where a use case gets defined, an external team comes in to deliver it, and the project closes when the output is live.
The question nobody writes into the brief is this: two years from now, when priorities have shifted and new opportunities have emerged, what has genuinely changed about the organisation's ability to work with its own data and act on its own intelligence?
In most engagements, that question goes unasked. It's worth asking.
There are two fundamentally different ways to run an AI programme, and most organisations arrive at one of them without ever consciously choosing.
In the first, the external partner is the intelligence, expertise gets bought, problems get solved, and the relationship continues for as long as the problems do. It can produce solid results, it also tends to create a specific kind of fragility - one that only becomes visible when something needs to change and the organisation realises it can't change it without going back to the same partner.
In the second, the partner is a catalyst. Knowledge transfers alongside delivery, the engagement has a natural arc toward independence rather than ongoing reliance. By the end of the programme, the organisation owns something, not just the output, but the understanding behind it.
The organisations consistently pulling ahead on AI are, in the main, in the second camp. They're building something that compounds, not something that requires continuous external investment just to stand still.
When AI capability sits primarily with an external partner, a few things become quietly difficult. Adapting when priorities shift requires understanding the logic of what was built, not just access to its outputs. Diagnosing underperformance requires practitioners who know the data, deciding where to invest next requires the kind of institutional knowledge that only comes from doing the work.
The dependency isn't just financial, it's strategic. The organisation becomes a consumer of intelligence it doesn't fully understand, which makes it harder to hold partners accountable, harder to spot when the approach needs to change, and harder to build on what already exists.
McKinsey's State of AI 2025 found that while 88% of organisations now use AI in at least one function, only 6% qualify as genuine high performers, where AI makes a measurable difference to the bottom line. The gap between investment and impact isn't primarily a technology problem, a significant part of it is an ownership problem.
None of this is an argument against external expertise, in the early stages of a programme, when the internal capability isn't yet there to make the right architectural and sequencing decisions, good partners are often essential. The question is what the engagement is building, not just what it's delivering.
At TXP, this is the distinction we try to be explicit about from the start. The work we do is designed to leave something behind: a team that understands what was built and can extend it, data foundations the organisation owns outright, and internal practitioners who've worked closely enough alongside us to make the next decision without starting from scratch. The engagements that progress fastest aren't always the ones with the largest initial investment, they're the ones where internal capability grows alongside external delivery, so each phase is owned more deeply than the one before.
Ask the partner to explain what your organisation will be able to do independently in 18 months that it can't do today.
If the answer is largely about using the output, that's useful information as it tells you clearly what the engagement is, and isn't, structured to deliver. That might still be the right choice.
If the answer includes understanding the data behind the output, extending what's been built as the business evolves, and applying the same logic to adjacent problems without needing the same level of external support - that's a different kind of engagement. One where the investment compounds rather than just accumulate.
A good partner will answer that question clearly and show you how the engagement is structured to deliver on it. In our experience, it's the question most organisations wish they'd asked earlier.
The whitepaper sets out a practical framework for structuring this - and what to look for in a partnership built to leave something behind.