Alongside everyday AI, place one big bet that could move revenue or profit

Alongside the everyday use of AI, I want a company to take one big bet at a time: something that, done correctly, can deliver new revenue into the business or significantly cut costs in a core process. Focus matters.
The biggest problem I see in AI initiatives is entropy: effort is spread very thinly across the whole organisation. Pilots formed without executive backing stall at the technology level or inside departments.
Independent analysis shows that 43% of CEOs see more revenue or lower cost from their AI work, while almost the same again say their companies are stuck. The difference is in the choices leaders make about where to put effort, and in whether they measure what happens.
Everyday AI will not show in your top or bottom line
I see repeatedly that organisations roll out a desktop AI tool, Claude or Copilot for example, to large numbers of their people. Everybody needs an AI tool to help them along. None of that is going to show up in the top or bottom line of the business, though, because it is all absorbed into day-to-day work.
A survey of our clients showed that, over a normal day, the use of AI tooling saved an average individual something like half an hour. Thirty minutes a day is 2.5 hours a week, so with 100 people you have saved 250 hours a week, which is six heads or more. Across a medium-sized organisation that is impressive, but you cannot consolidate those hours, and you are not going to remove headcount that way. All of it is important, and none of it changes the metrics in the business.
What a big bet looks like
Once you have organised and understood your AI use cases and ranked them against return on investment and risk, look at taking one big bet. It might have a fair amount of risk associated with it: you could spend quite a lot of money and it would not work, or done poorly it could damage your brand, your reputation or your service.
A classic example is the whole pipeline from proposal to statement of work in a services business. It can take multiple person days to get a decent mid-six-figure proposal out of the door. The vast majority of it can be automated using AI pipelines, and we have demonstrated that both for ourselves and for some of our customers. It cuts out lost sales time and returns engineers from pre-sales work to the rest of the organisation. Another good example is a prediction model, developed on your own data, that improves a service you offer to your customers and, done correctly, can deliver an entire new service and revenue line.
We delivered an AI capability to a consulting organisation that allows its customers to "talk to their data" rather than having solely to consume full reports. It has been instrumental in turning the company from a consulting organisation into a data organisation, with significant positive impact on its revenue and valuation.
Why fewer bets, led by the CEO
If you poorly fund many initiatives across the business, you end up with competing demands back onto the technology team, fragmented datasets and a fragmented platform for AI governance and delivery. A large number of those projects are likely to wither on the vine.
A larger initiative, one that moves a business process from primarily human to primarily AI, may go through some tough times before it comes out the other side. To get through that, the business needs to be clear about what it is trying to achieve and how it is going to fund it, and not leave it to be lost among other initiatives. That is why, if you fund fewer large-scale initiatives better, more of them will succeed over time.
So these bets should come from the CEO downwards, be funded appropriately and be pushed to happen in a short to medium timeframe. In my experience CEOs are very good at working out risk and reward in strategic initiatives, and generally only the CEO has the broader view of how the organisation itself can be transformed.
I have had both ends of the spectrum when putting the one big bet idea to leadership teams. Some have accepted the principle with enthusiasm. Others have been too stuck in the weeds to see what they could cull from their existing programme, and I can see that they are being less effective over time because their attention is too distributed and diluted.
Know what it costs to be wrong
A bet is started from a baseline and a goal, funded in stages behind stop points, and either doubled down on or stopped at a known cost. There are typically two stop points, with criteria agreed with the sponsor before work starts. One client's plan from a recent engagement to develop a prediction model shows the shape well: an investigation planned over ten weeks, as the cheapest credible test of whether the client's own data can show what determines the result on its core process. The first stop point was set at the end of week three and asked whether the data was good enough to go on. If not, the work would stop there. The second was a few weeks later, following deeper investigation, and asked whether a model could be developed that could show high enough accuracy to be of genuine value. The client's own criteria decide, and its experts keep the final call. The last two weeks were put aside for a report of what was found, either way, with a costed next step only deliverable should the model have proven its capability. Finding that the idea does not work was, itself, still a successful result, at a known price.
We worked with an organisation that wanted to cut the headcount in one of its primary functions by 10, from 50. They initially suggested just to build a proof of concept and demonstrate it working. We advised them against this, and instead helped them with a study of what was required: interviews across the staff and a time and motion analysis, to understand how and where headcount could genuinely be reduced.
We showed that they were likely to reduce headcount by three or four at most. The process load, although high, was broadly spread across a number of domain experts, and none of its elements could be reduced significantly enough to make the reduction they needed. The study was not expensive. They were not able to do the big bang they were hoping for, but the study has given them a path forwards for automating elements of that process and, over time, seeking the savings they are looking for.
What I say to a sceptical deal partner
Private equity owners need to see value in initiatives like these. A big bet pushed from the top is one of the things that can demonstrate to them, and to other stakeholders, that AI maturity is really developing in the business. That can be critical when it comes to exit preparations, in helping a company show that it is moving forwards in an AI world and can create new value.
A sceptical deal partner is going to say that there is a lot of risk involved in this, and ask why not go with something safer. It is a risk-reward balance. A more heavily regulated organisation may decide that its big bet is going to be lower down the risk scale, and accept lower reward in the process. One with a significant AI disruption looming ahead of it may need to take higher-risk, higher-reward opportunities to seize the advantages that AI offers.
Ultimately, the stakeholders need to decide what risk they are willing to accept, and what downside they are willing to accept if they do not take it. One of the strongest ways to mitigate the risk is to take on a dedicated AI partner to aid you in the transformation.
What an operating partner should ask for
Attercop does a lot of AI tech due diligence, so we get to see what good and bad AI strategy looks like in a business. I expect operating partners to ask every one of their portfolio companies for a clear AI strategy, and I expect those companies to be able to deliver it as a written plan. The plan should come with an 18-month to 24-month roadmap of actions and initiatives, set against a vision of where the company wants to end up. For almost all organisations that vision should involve at least some element of transformation, whether just in certain back-office functions or core across the entire business.
If a leadership team cannot express this succinctly and effectively to its operating partner, whether during a deal or in the middle of a hold, it is deficient and should be taken to task for it. Similarly, I would expect every portfolio company to have an AI disruption analysis to hand, which outlines its thesis on how AI is going to affect its market, and the company specifically, over the next five years. That is a very long window for AI, but without that forward plan you cannot say that the business strategy is in strong shape.


