Global enterprises are projected to spend $2.6 trillion on AI in 2026 and 2027. Around 60% of them will see minimal or zero material return on that investment. By rough calculation, that is close to a trillion dollars going to waste.
The goal is not to debate whether AI is a bubble. It is not. The goal is to understand why so much capital is being misallocated, and what the companies getting real returns are doing differently.
Episode 6 of Optimise covers the optimal AI capital allocation framework for business owners and CFOs who want to stop pouring money into AI ineffectively and start generating measurable ROI.
The false choice between buy and build
Most financial decision-makers approach AI the same way they approach other software: buy an off-the-shelf solution or build something in-house. With AI, neither option works well on its own.
Buying off-the-shelf AI licences creates tool sprawl. You end up with a collection of SaaS subscriptions, many running the same underlying model your competitors use, that function as enhanced search engines rather than strategic assets. You have no intellectual property. No competitive advantage. Just recurring costs.
Building entirely in-house looks like the alternative, but the numbers tell a different story. Research shows 67% of in-house AI builds fail, for three consistent reasons. First, retaining specialised AI engineers, particularly AI security engineers, as full-time employees is prohibitively expensive for non-tech enterprises. Second, foundational models update rapidly, meaning custom code built around them becomes obsolete quickly and requires constant refactoring. Third, internal developers end up spending most of their time connecting data sources and maintaining infrastructure rather than building the business logic that actually matters.
Spending a million euros building something in-house that is already outdated by the time it ships is not a strategy. It is a sunk cost.
The co-build model
The answer sits between the two extremes. Partnering with specialist architects to build custom enterprise AI infrastructure, where you retain full IP ownership but benefit from a team that does this work every day, consistently outperforms both alternatives.
The timeline for a co-build engagement typically runs four to six months, compared to six to eighteen months for a full in-house build. The success rate is higher because the partner team is not attempting this for the first time. And unlike off-the-shelf solutions, the intellectual property belongs entirely to you.
A 2026 Gartner study on enterprise AI budget benchmarks puts numbers to this. In-house builds return around 1.2x ROI. Off-the-shelf SaaS returns around 1.8x. Co-build partnerships return around 4.2x. The gap between buying licences and co-building with a strategic partner is not marginal. It is more than double.
What smart capital allocation actually looks like
The practical framework the episode outlines has three steps.
- First, audit your current tool sprawl. Identify every AI subscription your organisation is running and what it is actually producing in measurable output terms.
- Second, identify your high-value friction points. These are the operational bottlenecks costing the most in time or money that a custom AI solution could eliminate. Not every process needs AI. The ones that do should be prioritised by financial impact.
- Third, co-build targeted solutions for those friction points through a strategic engineering partner. Custom middleware built around your specific business logic, owned by you, integrated into your existing infrastructure.
This is also where AI-enabled software development differs from simply adding AI tools to an existing stack. The distinction between a collection of licences and a coherent AI architecture is where ROI is won or lost.
The underlying principle
AI investment decisions are not IT decisions. They are capital allocation decisions, and they deserve the same rigour. The companies heading into 2027 with strong margins will not be the ones that bought the most licences. They will be the ones that identified the right friction points and built targeted solutions around them with partners who specialise in exactly that work.
Watch episode 6 of Optimise here.
If you are trying to move from tool sprawl to a coherent AI investment strategy, that is the conversation Cleverbit is built for. Book a call to talk through where your capital is going and where it should be.