Part 3 of a four-part series on integration, value destruction, AI, and leadership
In the first two blogs in this series, I described why I have been spending time with firms and teams several years after their acquisitions, and how those conversations reveal a surprisingly binary outcome. Either the deal worked and people are broadly happy, or it didn’t and the problems are still visible long after integration was meant to be complete.
One of the most consistent accelerants of both success and failure in those conversations is artificial intelligence.
AI is no longer an abstract capability or a future roadmap item. In many of the acquired businesses I have spoken to, it is already embedded in how work gets done every day. And increasingly, it is the smaller company that is further ahead.
The unexpected asymmetry in AI maturity
Most large enterprises now have formal AI programmes. They have centres of excellence, governance frameworks, ethical guidelines, and approved toolsets. On paper, this suggests that AI integration should be straightforward.
In practice, we are seeing the opposite.
Smaller businesses, particularly digitally mature SMBs, often use AI in far more practical and deeply embedded ways:
- Automating large parts of back-office operations such as order to cash
- Reducing or removing layers of manual coordination
- Using AI-assisted forecasting and planning at a level of granularity enterprises often lack
- Orchestrating work through chat-based operating models, typically using Slack, where systems are integrated end to end
This creates a form of operational leverage that is easy to underestimate during diligence, and very hard to dismantle without consequence.
Why AI makes integration structurally harder
Historically, integrating a small business was largely a systems exercise. Migrate finance, align CRM, standardise reporting, and move on.
AI breaks that model.
When AI is embedded into workflows rather than bolted onto systems, the business becomes tightly interwoven. Tools, data, decision-making, and communication reinforce each other. Removing or replacing one element affects everything else.
Several corp dev teams described this as a “cat’s cradle” business. Pull one thread and the entire structure moves.
Faced with this complexity, a common response has emerged.
Leave the business alone for eighteen to twenty-four months.
The intention is sensible. Reduce risk. Buy time. Let the dust settle.
The outcome is usually the opposite.
AI continues to evolve inside the target. Automations deepen. Dependencies increase. And the gap between the target’s operating model and the enterprise environment grows wider, not narrower.
By the time integration resumes, the cost, risk, and disruption are materially higher than they would have been earlier.
Why this is more than an ethics or policy issue
There are, of course, necessary AI considerations that every acquirer must address:
- Data provenance and security
- Ethical use and regulatory compliance
- Model transparency and explainability
- Alignment with corporate policies
These are important. But they are not the main problem.
The deeper challenge is economic and operational.
In several post-acquisition conversations, it became clear that the AI capabilities underpinning the target’s efficiency would be extremely expensive, or operationally impractical, to replicate at enterprise scale. Standardising them away destroyed speed. Replacing them removed insight. Deferring decisions increased dependency.
As discussed in Blog 2, this kind of regression rarely shows up immediately in the P&L. It shows up in slower decisions, more people required to do the same work, and declining morale.
What the research is starting to show
This lived experience is increasingly reflected in external research.
McKinsey & Company has shown that a majority of AI transformations fail to capture their expected value, most commonly because operating models and decision processes are not redesigned alongside the technology. Their research consistently highlights that embedding AI into daily workflows is significantly harder than deploying tools in isolation.
Gartner has reported that organisations integrating AI-intensive capabilities experience longer integration timelines and higher complexity, particularly where AI is deeply embedded in workflows rather than contained within discrete systems.
Harvard Business Review has also noted that smaller firms often generate higher returns on AI investment due to focus, speed, and tight integration into operations, advantages that are frequently lost post-acquisition when standardisation overrides capability preservation.
What is often missing from these analyses is the long view. How these AI decisions feel to teams several years later, once the tools have settled and the integration choices are no longer reversible.
Four fixes corp dev teams need to own
Across integrations where AI complexity was handled well, four patterns stand out.
1. AI discovery focused on value creation, not just risk
AI diligence must go beyond tool lists and policies.
Corp dev teams need to understand:
- Where AI genuinely removes work
- Which decisions depend on AI-generated insight
- What operational advantage would be lost if those capabilities disappeared
Without this, integration decisions are blind.
2. An explicit AI integration roadmap, not deferral
Leaving AI-heavy businesses untouched is still a decision, and often a costly one.
Successful acquirers explicitly decide:
- Which AI capabilities remain standalone
- Which integrate with enterprise platforms
- Which are deliberately scaled across the group
This design work must happen early, not as remediation.
3. Protect AI-driven advantage as a value metric
Most integrations track cost synergies and risk reduction. Very few track what AI actually delivered pre-deal.
Speed, automation coverage, and decision quality must be measured and protected, or they will be lost.
4. Be explicit about AI leadership and honest about where expertise lives
One of the most overlooked decisions is who leads AI post-acquisition.
In many cases, the most practical AI knowledge sits with the acquired company, not the acquirer. Successful integrations recognise this and allow leadership to flow toward expertise, at least initially.
Where authority and knowledge are misaligned, AI becomes a source of friction rather than advantage.
How this connects to the rest of the series
AI amplifies everything discussed in Blog 2. Poor integration design destroys value faster. Good design preserves advantage longer.
But AI also exposes a deeper issue, which is the focus of the final blog in this series.
Leadership.
In Blog 4, I will explore why the difference between integrations that create careers and those that create quiet failure almost always comes down to leadership decisions made early and left unchallenged.
AI does not remove the need for leadership.
It makes the cost of poor leadership higher.
References
- McKinsey & Company – The state of AI in 2023
https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
