The AI leapfrog effect: Why standardizing now could be your costliest mistake

Towards the end of 2022, something shifted in enterprise boardrooms. Convinced by the promise of a genuinely new class of technology, businesses began investing in AI across marketing, sales, developer platforms, analytics and automation. The experimentation was justified. Use cases were unclear, the cost of not participating felt higher than the cost of overspending, and no one wanted to be the organization that missed the wave.

Today, however, the climate is changing. The AI market is maturing as the first wave of experimentation comes to an end. Now, enterprise leaders are putting their AI systems under review and asking: “Are the outcomes worth the price?”

According to one recent study, actual revenue growth and profitability improvement are the primary ROI metrics for AI projects (selected by 22% of respondents), ahead of productivity (18%). Executives today care less about narratives that focus on efficiency or time savings and more about the impact of AI on the bottom line. To some extent, this is a consequence of just how heavily some organizations have invested in the technology. Some businesses support AI portfolios worth up to five points of EBITDA, (figures from BCG put the average spend this year at 1.7% of revenues). At that level of investment, AI is no longer an innovation budget line. It is a balance sheet exposure, and boards are starting to treat it accordingly. Given how much is being spent, the pressure to demonstrate returns is building, and the pressure to demonstrate governance over that spending is not far behind.

Over the next 12-18 months, we can expect to see something of a reckoning. Vendors with products that deliver only marginal productivity gains can expect to lose out to competitors that offer better value for money. This phase of market consolidation will likely be significant and it’s possible that many of the AI companies that attracted serious enterprise contracts over the last two years will not survive.

The AltaVista trap

What makes this moment particularly difficult to navigate is a dynamic I think of as the AI Leapfrog Effect, and it is unlike anything enterprise IT leaders have encountered before.

The AI market is not evolving incrementally. Platforms are overtaking one another in capability, cost-efficiency and architectural approach at a pace that enterprise procurement cycles were never designed to track. The leading tool in any given category today may occupy a distant second position within 90 days. And critically, the leaps are not marginal. They are structural shifts in underlying model capability, reasoning depth or integration architecture that render yesterday’s best answer genuinely obsolete.

History offers a useful warning. In the early days of the internet, AltaVista was one of the leading web browsers. It had an early-mover advantage and, in the early days at least, a strong user base. Yet, as we all know, AltaVista lost out to Google, a new kid on the block that had the advantage of much better technology. Google won because it learned where the pioneers were going wrong and was able to overtake them through rapid technological advancements.

Today, many of the current AI leaders risk being leapfrogged in the same way. Indeed, the environment today is much more perilous to first-wave pioneers than it was back in the mid ‘90s. AI technology is evolving much faster than any technology that has come before. A leading platform in any given category today may slide into distant second in as little as 90 days.

For enterprise IT users, this creates a strategic trap. The AI market is moving much faster than procurement cycles can track. A standard enterprise software evaluation (RFP, shortlist, proof of concept, legal review, contract) can easily consume six to nine months. In AI terms, that is two or three capability generations. Businesses that standardize too early risk being locked into a platform that’s soon out of date. Conversely, if organizations delay key decisions too early, they may fail to capture the productivity gains that competitors are already realising.

The best course of action is to track close enough to the market to understand its direction of travel without over-committing. Practically speaking, this requires carrying out regular technology evaluations while keeping options open across two or three platforms. It’s also important to resist pressure from vendors to sign long-term contracts for platforms that are still in development or to undertake complex integrations. Vendors pressing hardest for long-term commitment are often the ones with the most to lose if you wait.

Striking while the iron’s hot

The one major caveat to all this is that in a few niche areas a clear leader will have already emerged – the Google to the AltaVistas of this world. In these cases, the gap between the leaders and the rest is so large it’s unlikely to be closed anytime soon. Here, businesses can invest with much greater confidence.

Developer tooling is the clearest example. At Avantra, we evaluated every serious option available for AI-assisted coding, including some of the less widely discussed platforms. Claude Code outperformed everything else we tested, and the margin was not close enough to generate any real debate. For our developers, Claude Code is so dominant there’s no point in maintaining Gemini or OpenAI licences, and we’ve therefore been able to benefit from cost consolidation.

That kind of clarity is useful, and organizations should act on it where they find it. The mistake is assuming that because one category has settled, the others have too. In areas like AI-assisted design, marketing content generation and enterprise search, the picture remains genuinely unsettled and standardising now carries real risk.

What consolidation will look like

The AI tools that will survive the coming reckoning share a few characteristics. First, they’re deeply embedded in workflows rather than sitting alongside them. Second, they deliver measurable, attributable outcomes rather than general productivity improvements that are hard to isolate. And third, they serve use cases where the switching cost is high enough that users do not migrate the moment a competitor releases an update.

Tools that fail those tests, particularly the ones that gained traction because they had no good alternative in 2023, rather than because they were genuinely superior, will face a sharp correction. Some will be acquired for their user bases or technical teams. Others will simply lose enterprise renewals at a rate that’s not sustainable.

Organizations best placed to navigate this period of market turbulence will have maintained honest internal records of what value their AI investments deliver. Rather than being measured by anecdotal evidence from enthusiastic early adopters, success will be quantifiable through gains in output, quality or cost at a team or process level. That discipline is harder to maintain during the experimentation phase, but it is what separates the organizations that will make smart consolidation decisions from those that will simply cut platforms indiscriminately when the pressure arrives.

What to do now

As organizations review their AI investment strategy there are three fundamental steps to take:

1. Resist multi-year platform agreements in categories where the technology is still visibly moving. Negotiate annual terms where possible and insist on contractual clarity around capability benchmarks and data portability. If a vendor will not agree to data export provisions, treat that as a significant red flag.
2. Document what your AI tools are producing in terms that will survive a CFO’s scrutiny. This means outcome metrics tracked at the team or process level, not survey-based satisfaction scores or anecdotal reports from enthusiastic early adopters. If you cannot produce this data today, instrument for it now, before the renewal conversation begins.
3. Move decisively in categories where consolidation has already happened. Where a clear leader has emerged, and the capability gap is structural rather than marginal, consolidating onto that platform and removing competing licences is straightforward value capture. The cost savings fund the optionality you need to maintain elsewhere.

When your board asks what the company is spending on AI, make sure you walk in with two numbers: what you are spending, and what you are getting. The organizations that cannot produce the second number are the ones that will be cutting indiscriminately in 18 months. The ones that can will be compounding the advantage they have already built.

The AI Leapfrog Effect is not a reason for paralysis. It is a call for precision: know where the market has settled, know where it has not and build your investment strategy around that distinction rather than around vendor timelines or the anxiety of being left behind. It means resisting the impulse to treat AI procurement like software procurement, where standardization and consolidation are almost always virtues. In AI, right now, selective standardization is a virtue, and premature standardization is a liability.

The organizations that internalize this will not just survive the consolidation wave but define the competitive landscape on the other side of it.