Every SaaS company is shipping AI features. Far fewer are reasoning well about the AI strategy.
The board is asking what your AI strategy is and whether you are behind. Shipping features answers neither. The hard part is not putting AI in the product. It is reasoning well about where it actually defends the business.
Request a Strategic Briefing →Ask any software leadership team what they are doing about AI and they will point at the roadmap: the copilot, the assistant, the agent shipping next quarter. That is the product, and the pressure to ship it is real. The trouble is that the product is not the strategy. Shipping AI features says nothing about where AI defends the moat, what is worth building versus buying, what not to build at all, and which platform and pricing bets to make. Those calls need judgement under fast change, and a commodity AI hands back a fluent, confident answer to every one of them without ever pushing back. Gildoni installs the Havruta Methodology (formerly the Think Partner Methodology) into how technology leadership reasons with AI, so the strategy is as sharp as the feature set. This is not an AI platform. It is the reasoning discipline above the roadmap.
The whole sector is racing to ship AI, and most of it is not paying off
This is not a niche anxiety. It is the defining pressure on every software company right now: ship AI into the product, bolt it into the roadmap, and answer the board when it asks whether you are behind.
The adoption curve backs that up. Industry has become the engine of AI, and the rush to embed it into software is no longer optional. But the gap between shipping AI and earning a return from it has become the real story of the past year, and the people closest to it are clear that the failure is one of method, not of model.
The numbers show the race, and the gap inside it.
of organisations reported using AI in at least one business function in 2024, up from 55% the year before. Generative AI use jumped from 33% to 71% in a single year.
of organisations are seeing no measurable business return on generative AI, while just 5% of integrated pilots are extracting real value. The report names it a learning and method gap, not a model one.
of enterprise applications will integrate task-specific AI agents by the end of 2026, up from fewer than 5% in 2025. The feature race is accelerating, not slowing.
The race is real. The return is not following it.
Generic tools like ChatGPT excel for individuals because of their flexibility, but they stall in enterprise use since they don't learn from or adapt to workflows.
Why shipping features gets mistaken for having a strategy
If the whole sector is shipping AI, why is so little of it paying off? Three reasons.
The demo is visible and the judgement is not. A shipped feature can be shown to a board in ten minutes. The reasoning behind where AI defends the moat, what not to build, which bet to make, cannot. So the visible thing stands in for the invisible one, and the roadmap becomes the answer to a question it never asked.
The per-task question is the wrong question. Teams burn cycles asking "can AI do this feature?" The answer is almost always yes, which is exactly why it tells you nothing. The hard question is which features defend the business and which are commodity the moment a model provider ships them natively. That is a judgement call, and judgement is what gets skipped.
And the speed that makes AI valuable is the speed that outruns the thinking. The MIT evidence is blunt: most enterprise generative AI shows no measurable return, and the cause is method, not model. The strategy is being made at demo speed while the market moves at platform speed.
It is worth being honest about the other side of this. The disciplines of good product strategy, knowing your moat, choosing what not to build, deciding under uncertainty, are old. AI has not rewritten them. It has stress-tested them, and found most leadership teams answering strategy questions with a feature list.
Shipping AI features is the easy part
Here is the part the roadmap leaves out. Every competitor is shipping the same copilots, the same assistants, the same agents. The feature set converges fast, because the models are shared and the capability is buyable. What does not converge is the judgement underneath it: where AI actually defends the moat, what to build versus buy, what to leave alone, which platform and pricing bets compound.
That is the real gap, and it is not a build gap. Put a platform bet or a build-versus-buy question to a commodity AI and it will hand back a fluent, confident answer without ever asking what you have missed. At the altitude where the decisions are largest, that is the Mirror Principle at its most expensive: if the reasoning going in was generic, the strategy coming out is generic, however polished the deck reads. A model can build any feature you specify. It cannot make the strategy that decides which features matter any good. That is a different discipline.
AI enters low, shipped as a feature, which is where most software companies stop. Its real weight rises to the leadership level that owns the moat, the platform bets and the decision about what not to build. The red is the distance between where AI is shipped and where the strategy is actually decided: built, but not reasoned through.
What the Havruta Methodology installs at leadership level
The Havruta Methodology is that discipline. It changes the default behaviour of the machine from agreeing with you to reasoning with you, which is precisely what a platform bet or a build-versus-buy call needs.
The Flip
The Flip puts the machine on the other side of the question. Instead of confirming the roadmap, it argues against it: where is this feature commodity the day a model provider ships it natively, what bet are we not making, what would have to be true for build to beat buy. The leadership team gets challenged before the market does the challenging.
Ground Truth
Ground Truth keeps the reasoning anchored in your actual position, your real moat, your data advantage, your cost base and your customers, rather than in the generic product strategy an AI produces by default. A platform bet built on a plausible average is worse than no AI at all.
Decision Velocity
And Decision Velocity lets the team decide at the speed the model layer is moving, compressing the path from question to defensible bet without surrendering the judgement to the machine.
The fuller account of how all of this works is on the methodology page.
What this is not
This is not an AI platform and it is not a dev tool. It is not an MLOps stack, a model API, a copilot you embed, or a feature you ship. It is not AI training or general AI literacy. The tooling and the infrastructure are a separate market. This is the thinking above them.
It changes how the leadership team reasons about the AI strategy it already owns: the moat question, the build-versus-buy call, the what-not-to-build decision, the platform and pricing bet.
Where a leadership team starts
The methodology is installed along a ladder, and a leadership team enters at the rung that fits.
Most begin with the Eye-Opener Workshop, a half-day in which the team sees the shift on its own real strategy work.
A leadership group embeds the practice through the Havruta programme, taking the discipline across the team.
A single high-stakes question, a platform bet, a build-versus-buy call, a what-not-to-build decision, can be worked through Advisory Havruta.
How a CIO and technology leadership reason with AI
For the technology leaders specifically, the role page takes the same discipline to the function that owns the stack and the build decisions day to day. A Strategic Briefing is how to decide where to begin.
Go to the CIO pageLeadership questions about AI in software companies
Is shipping AI features the same as having an AI strategy?
No. Shipping AI features is the product. The strategy is the judgement underneath it: where AI actually defends your moat, what is worth building versus buying, what you should not build at all, and which platform and pricing bets compound. Those calls are invisible in a demo, so the visible feature set ends up standing in for them. When the board asks what your AI strategy is, a roadmap of copilots and agents answers a question it never asked. The strategy is a separate, harder discipline.
Why do most enterprise AI initiatives show no return?
The MIT Project NANDA research found the large majority of organisations seeing no measurable return on generative AI, with only a small fraction extracting real value. The authors are clear it is a method and learning gap, not a model one: the models are capable, but the way they are deployed and reasoned with is not. For a software company that means the answer is rarely a better model. It is sharper judgement about where AI is worth applying and how the leadership team reasons through that decision.
How should a software company decide build versus buy on AI?
Start by dropping the per-feature question of whether AI can do it, because the answer is almost always yes and it tells you nothing. The real question is which features defend the business and which become commodity the moment a model provider ships them natively. That is a judgement call about your moat, your data advantage and your cost base, not a capability check. The discipline is to make the AI argue against your own answer, anchor it in your real position, and decide at the speed the model layer is moving.
Are we behind on AI if competitors are shipping more features?
Not necessarily. Feature sets converge fast because the models are shared and the capability is buyable, so shipping more of the same is rarely a durable lead. The thing that does not converge is the judgement underneath: where AI defends your moat, what you choose not to build, which bets compound. Being behind on features is recoverable. Being behind on the reasoning that decides which features matter is the gap that actually costs you, and it is the one a longer roadmap will not close.
Is this an AI platform, a dev tool or training?
None of those. It is not an AI platform, an MLOps stack, a model API, a copilot you embed, or a dev tool, and it does not touch your codebase. It is also not AI training or general literacy. Those address the build and the skills. This addresses the thinking: how technology leadership reasons through a strategy decision so the answer is genuinely theirs, anchored in their real position, and stress-tested before the market does it for them. The tooling is a separate market. This is the discipline above it.
Where should we start?
With a Strategic Briefing, or with the Eye-Opener Workshop, where a leadership team sees the difference between instructing AI and reasoning with it on its own real strategy work. From there the path depends on whether you are setting the AI strategy at leadership level, embedding the practice across a leadership group, or working a single high-stakes question such as a platform bet or a build-versus-buy call.