How To Choose An AI Consultancy: The Five Questions We Ask Before Taking On A Client

How do you choose the right AI consultancy?

You choose an AI consultancy the same way you'd choose anyone you're about to hand a chunk of your leadership team's time and a meaningful budget: by finding out, before either of you signs anything, whether you are genuinely ready for what you're buying. After two decades running transformation programmes inside banks, insurers and manufacturers, I can usually tell within the first hour of a conversation whether an engagement is going to land or quietly stall.

The tell is rarely in the plan. It's in the room. Most of the consulting industry's business model depends on not asking these questions too carefully. A fuller pipeline looks better in the quarterly numbers than an honest one, so the incentive, across the industry and ours too if we're not careful, is to keep every prospect moving forward regardless of fit. I think that's a bad trade dressed up as commercial sense. The client with the wrong fit spends months in workshops that produce good-looking documents and no changed behaviour, spends real money doing it, and walks away having quietly concluded that AI transformation doesn't really work, when the truer conclusion is that this particular engagement, at this particular time, was never going to.

What questions should you ask an AI consultancy before you sign?

Ask yourself the same five questions we ask before we take on any client. In order: are you ready to act or still deciding whether to be interested, do you want a quick win or the whole transformation, are you willing to make the hard calls, is the right person actually in the room, and can you say specifically why now. None of them are about budget or technology: all of them are about whether the conditions exist for the work to survive contact with your actual organisation.

I'll take each in turn, because in twenty years of watching engagements either compound or evaporate, the honest answer to each has told me more than any proposal document ever has.

Are you ready to act, or still deciding whether you're interested?

The tell isn't enthusiasm: some of the clients we've done our best work with sounded almost sceptical on the first call. The real tell is whether the decision to act has already been taken, or whether you're still evaluating from a safe distance and calling it due diligence.

I spent a lot of years sitting in steering committees where "we're exploring options" was the standing agenda item for the third quarter running. Nobody was lying. Everyone genuinely intended to get to a decision eventually. But intention without a decision is just a well-organised form of waiting, and waiting doesn't show up as a risk on anyone's dashboard until eighteen months later, when the business case has gone stale and the market has moved on without you. Contrast that with the leadership teams who've already cleared a manager's diary and reallocated a slice of this quarter's budget before the contract is even drafted. You can hear the difference down the phone. One is describing a decision they've made. The other is describing a decision they're hoping the call will help them avoid making.

Do you want a quick win, or the whole transformation?

Both are legitimate answers, provided you name the one you actually mean. What isn't legitimate is calling a two-day workshop "transformation" and being disappointed a year later that the way a hundred people work every Tuesday morning hasn't changed.

I watched this pattern for years inside larger organisations, usually dressed up as an innovation lab. A capable team builds something genuinely impressive, the executive committee applauds it in a showcase, and then it sits exactly where it was built because nobody ever scoped it as more than a demonstration. That isn't a technology failure. It's a scoping failure, and it's entirely avoidable if you decide upfront which of the two things you're actually paying for: a bounded, honestly priced piece of work that fixes one thing well, or a genuine change to how the business runs, which costs more, takes longer, and touches more people than a workshop ever will. Both are fine. Confusing one for the other is what wastes the year.

Are you willing to make the hard calls a transformation demands?

Every transformation worth the name eventually asks you to choose the outcome over something you'd rather leave alone: a process that has outlived the reason it was built, a tool a team has grown genuinely attached to, occasionally the shape of a role that no longer needs doing the way it's being done. A leader who wants the upside without ever sitting through that discomfort is asking for something that doesn't exist.

I want to be precise about what I mean here, because this is the point where people hear "job losses" and switch off, and that isn't what I'm describing. The whole premise of this work is a business that becomes five to ten times as capable with the same people already in it, not the same output with fewer of them. But real capability doesn't arrive without real decisions along the way. I've sat with finance teams running a parallel spreadsheet for a decade because nobody trusted the official system a senior stakeholder had put their name to, and watched a transformation programme die not in one dramatic moment but in eighteen months of nobody being willing to have the conversation about which system was actually going to win. The businesses that get through this are the ones whose leaders have that conversation before it's forced on them, not after.

Is the person in the room the one who can actually say yes?

Usually not, and that's rarely a question of ability. Capable heads of operations, IT directors and transformation leads are, more often than you'd expect, the wrong primary contact: not because they can't do the work, but because they can't reprioritise a peer's time or approve spend that wasn't already in this year's budget without going back up the chain.

This is the pattern I learned to spot fastest in large organisations, because it's built into how they're structured. The sponsor shows up to the kick-off and the monthly steering update, nods in the right places, and the actual person who could unblock a stuck decision is three layers away and has never been in a room with anyone doing the work. I ask a version of one question early in every first conversation: who signs this off if it needs to move faster than planned. Watch who in the room glances sideways before answering. In a business your size, it's rarely three layers of committee; it's more often a managing director who has, with good intentions, stepped back from an area to focus elsewhere and installed a capable lieutenant with a title but not the authority that title implies. The engagement doesn't fail loudly when this is the case. It stalls quietly, a decision at a time, until you look up and a year has gone.

Can you say, specifically, why now?

Real engagements start from something concrete: growth that has outrun the systems supporting it, a first attempt at AI that's already been quietly shelved, pressure from a board or an investor, a near-miss where something almost went wrong in front of a customer. "We should probably be doing something with AI" isn't disqualifying on its own, but if there's nothing underneath it when I push gently, that's useful information about timing, not about whether you're a serious business.

In the large organisations I spent most of my career inside, the "why now" was usually imposed from outside: a regulator, an activist shareholder, a new chief executive with a mandate to prove. In a business your size it's more often internal and more personal: a founder stretched thinner than the org chart admits, a competitor visibly moving faster, an operations lead who watched something nearly fail last month and doesn't want to feel that particular fear twice. Either is a legitimate reason. The absence of any reason at all just tells us the timing conversation needs to happen honestly, rather than being assumed.

What does it mean if you recognise yourself in three or four of these?

It's a genuinely useful signal in either direction. Recognising real readiness in most of them is worth acting on now, and recognising the opposite is just as valuable to know today, before either side spends months and a meaningful invoice finding it out the hard way.

I said at the start that the outcome is usually visible in the first conversation, if you know what to listen for. That isn't a sales technique. It's the same instinct that twenty years of watching change programmes either take hold or fade away eventually teaches anyone who pays attention: readiness is a fact about an organisation, not a feeling in a meeting room, and it's kinder to everyone involved to find that out before the invoices start, not after.