Vetting is not validation

Ronnie Hamilton, pre-sales director at Climb Channel Solutions, argues that the channel’s real advantage in artificial intelligence is proof, not speed

Every AI product I get shown looks impressive, and that is the problem.

The channel has spent the past few years getting good at one half of the AI challenge: working out which vendors are worth selling. Distributors are overwhelmed by the number of vendors available to them, and sorting the contenders from the pretenders has become a discipline in itself. That work is real, and we do it too. But it is the easy half. It establishes that a vendor is a real company with a channel programme behind it. It does not answer the question the customer cares about, which is whether the product will work for them.

That answer tends to come later, in the proof of concept, where a vendor’s demo meets the customer’s own data and where pre-sales either earns the recommendation or finds the reason not to make it. The gap between what a product demonstrates and what it deploys has always existed. AI has widened it sharply. The industry’s answer is to move faster, certain the window of opportunity is closing. But this approach risks turning that gap into a bad recommendation..

A deadline the channel did not set

The channel is working to a deadline. When cloud arrived, partners had three to five years to retrain, hire and rebuild how they sold around it. Futurum puts the equivalent window for AI at 12 to 18 months, because the technology is now changing by the week.

Customers are applying their own version of the same pressure. In research we ran across 600 enterprise decision-makers in the UK, Ireland and Germany, 32% expect a return on AI within one to six months. And the channel is strangely absent while this is going on: 47% of those organisations are not engaging resellers or vendors on AI at all, even as 55% plan to expand their use of it.

The effect on a partner is clear. There is pressure to have an AI answer and pressure to have it now, against a backdrop in which every vendor’s deck leads with AI and every demo is impressive. The path of least resistance is to pick quickly and recommend quickly. It is also a trap.

Why AI resists judgement by demo

A demo runs on clean, curated data in conditions the vendor controls. A customer’s environment is none of those things. Their data is fragmented, access is governed, and edge cases are real.

The danger is that the product still installs perfectly well in that environment. The licence goes live, the pilot is written up as a success, and the failure is invisible at the point of sale. It only announces itself months later, when the customer asks what they actually got.

The scale of the mistake is already showing in the data. Over half of the organisations we surveyed (53.5%) place themselves in what we called the AI Gap: tools in use, but no value coming back. The confidence problem sits right alongside it. Some 61% describe AI as well understood and implemented in their organisation, while 49.2% have trained fewer than half their staff to use it. The belief that AI is working is running well ahead of the evidence that it is.

When that catches up with a customer, their disappointment falls on the partner who recommended the product, not the vendor who built it.

The discipline the market is skipping

So the answer is not to move faster. It is to validate before recommending, and validation is not the same discipline as vetting.

Vetting asks whether a vendor is sound. Validation asks whether this product does what it claims on this customer’s data, against this customer’s constraints, for the use case the customer actually has. That work happens in the proof of concept, it takes the time it takes, and it is the one stage that cannot be compressed.

It is also the stage a self-serve marketplace cannot perform, and the question a vendor, in the moment of a deal, is not placed to answer. It needs someone sitting between the two who has run the same product through enough real environments to know where it tends to break, and who will tell a partner a product is not right for their customer even when a quick yes would be easier.

This is the part of distribution that has nothing to do with moving stock. It is the judgement about whether a product fits, whether the use case holds and whether the risk is worth carrying, made for the long-term success of the deployment rather than the speed of the sale.

Vendors are not on the other side of this

None of it is at odds with what the vendor wants, either. No vendor is served by a sale that fails in production. A deployment that stalls is a renewal lost and a reputation dented in a market that talks.

The distributor that thinks past the transaction, and grounds what it carries in knowing where it works and where it does not, is looking after the partner, the customer and the vendor at the same time. A partner is better served by a team that tests a product honestly and sometimes comes back with a no rather than a fast yes that unravels in deployment.

Speed was never the channel’s advantage

None of this argues against the 12-to-18-month window. The window is real, and so is the demand behind it. The argument is about what the channel does inside it.

Vendors will always move faster than distribution, and marketplaces faster still. Speed was never going to be the channel’s sole advantage, and neither was the size of the catalogue. What the channel can still offer, when AI products demo better than they deploy and customers are in no position to tell the difference, is proof. The partners who build their AI practice on that principle will be the ones whose customers are still listening three years from now.

See more stories here.