Everyone’s still asking “what’s next for AI?” But the more important question might be: what happens when the excitement fades? Especially on a day when we’re being inundated with breathless excitement about new model releases. What’s the longer term arc?
Because markets often behave in familiar, predictable ways, even when the technology is new. If we know what phase we’re in, we can better anticipate what’s coming next.
I’m old enough to have seen this pattern before in the cybersecurity industry as it underwent its shift: over 200 companies were absorbed into just 11. Some - like Palo Alto and Cisco - now consist of dozens of acquisitions stitched together into a platform.
The takeaway isn’t just that markets consolidate. It’s that they often do so in predictable stages.
describes these as:Opening: New capabilities emerge. Innovation accelerates. Startups flood in.
Scale: The market crowds. Buyers push for integration. Platforms begin to form.
Focus: Consolidation accelerates. Strong players grow, weaker ones exit or merge.
Balance & Alliance: The space stabilises. A few players dominate. Attention shifts to trust, governance, and ecosystem cooperation.
What makes this framework useful is that it’s not unique to cybersecurity. It’s a pattern that plays out - in different ways, at different speeds - across most technology waves. And right now, AI, especially the LLM space, looks to be entering the same cycle.
The Opening phase came fast
From 2020 through 2023, we saw a full-blown Opening phase in LLMs. GPT-3 and then ChatGPT ignited a frenzy. Almost overnight, it felt like every tool and workflow was being rebuilt with an LLM at the centre.
Startups proliferated. Many looked similar. Differentiation was thin.
UX layers varied. Wrappers changed. But under the hood, the stack often pointed to the same few APIs.
And that was fine - for a while. In the Opening phase, novelty is enough. A good demo, a clever application, a thin layer on top of a foundation model - all of it could raise money, attract users, or win headlines. The question wasn’t whether the thing was sustainable - only whether it worked, or looked promising, or beat the market to it.
But markets evolve.
Scale: where we are now
Now, in 2025, the pattern is shifting.
Point solutions are giving way to platform expectations.
Buyers - especially in enterprise - are tired. Tired of evaluating redundant tools. Tired of fractured workflows. Tired of gluing together products that don’t talk to each other.
As in Ross’s second phase, the market wants coherence.
This is where incumbents have a structural advantage. Microsoft is embedding Copilot across its suite. Salesforce, Notion, Adobe - all integrating LLMs where users already live. Their advantage isn’t just the models - it’s proximity. Trust. Distribution.
Startups are adapting. Some are moving down-stack, into infrastructure and orchestration. Others are going deep into verticals where they can offer not just AI functionality, but domain-specific leverage. Some are pivoting entirely - away from product, toward tooling, training, or services.
But one thing is clear: we’re no longer in the novelty phase. We're in the fit-and-scale phase. That changes what gets funded, what gets adopted, and what survives.
Focus: what comes next
In the next phase - and we’re not far from it - the consolidation accelerates.
Not just through headline-grabbing M&A, though that will happen. More often, it will be quiet:
Startups fading from relevance as growth stalls.
Tools getting absorbed as features into broader platforms.
Infra players merging to stay alive.
Investment shifting to safer bets.
This phase demands defensibility.
Proprietary data. Distribution moats. Deep integration. Real value, not speculative buzz.
This is the phase where most of the noise clears. The market becomes more legible - to buyers, investors, and operators alike.
Balance & Alliance: the long game
This final phase may take years. But we can see its outline.
When it arrives, it will look like a few dominant ecosystems coexisting - perhaps uneasily. LLMs will be less visible. More embedded than advertised. More infrastructure than product.
Differentiation won’t just be in UX or raw model performance. It will be in interoperability, auditability, revocation, governance. Trust becomes structural. Control becomes architectural.
The frontier models may continue to evolve rapidly - but most organisations will be working within constraints: compliance, risk tolerance, supply chain, trust.
This phase doesn’t end the AI wave. But it hardens its structure. And in doing so, it narrows the aperture for experimentation at scale.
Strategic Implications: What This Means Now
This arc isn’t just academic. It has implications, especially for those building, buying, or investing in LLM-native products today.
1. Differentiation is shifting
In the Opening phase, a good demo could win. In Scale and Focus, it won’t. Differentiation now comes from:
Deep integration into workflows
Access to proprietary data or users
Domain fluency that generic tools lack
Novelty won’t carry you through the next phase. Specificity might.
2. Distribution will matter more than model quality
As commoditisation sets in, technical advantage alone won’t win.
Being embedded inside existing tools beats trying to replace them.
Owning the user relationship becomes more powerful than having a clever UI.
Startups that can’t find underserved distribution channels will struggle.
3. Value is moving down the stack
The attention has been on the surface - copilots, chat interfaces, summarisation layers.
But as usage grows, value moves deeper:
Orchestration
Observability
Trust and safety tooling
Memory and revocation
Fine-grained access control
It’s not glamorous - but it’s foundational.
4. Governance is no longer an afterthought
In the Balance & Alliance phase, governance isn’t just a whitepaper - it’s system design.
Can you audit the output?
Can you control who sees what?
Can you revoke memory?
Can you explain decisions?
These questions aren’t just for regulators. They’re already showing up in enterprise RFPs.
Final Thought
The pace of AI makes everything feel new. But structurally, this is a familiar pattern.
Markets open. Then they converge.
Novelty gets replaced by integration.
Excitement gets replaced by scrutiny.
And eventually, only a few players shape the long-term landscape.
We may not know how this will play out - but recognising the signs might help us make better bets along the way.
If you’re navigating these shifts, here’s how I partner with teams in the thick of it.”
Stuart x
Excellent article, Stuart👍The strategic implications you've illustrated are absolutely spot on! I believe all these 4 phases form a part of the overall Continuous Transformation cycle and taking a leap from from "Opening" phase to "Scale" phase is generally where most struggle with, and that's where the mindset to build for scale from the beginning comes into play.