When Big Companies Pull Back — It’s a Signal, Not a Victory

Lauren Herring recently noted that “big companies are starting to pull back on activity” – cutting back initiatives, postponing new launches, rethinking investments. The move is understandable: macro uncertainty, tightening budgets, and cautious boards all conspire. But the real risk is not that large firms pause — it’s how they pull back.

In my view, many of these retreats stem from a misunderstood lever: the interplay of AI and subject-matter expertise. I believe the next frontier isn’t simply “more AI” or “more business people” — it’s AI-first subject-matter experts who deeply grasp what’s happening underneath. Only then does the magic really happen.

Why the Pullback Is Hitting Now

Let’s start with the context:

Many AI or digital transformation efforts launched in the past 2–3 years were led by technologists or external consultants promising “AI will fix everything.”
But reality intervened — data silos, cultural resistance, unclear ROI, misaligned metrics.
When targets slip and budgets get tight, executives often cancel or deprioritize projects that seem speculative, especially those that don’t show immediate traction.

So, the big companies are making what looks like rational course corrections. But I worry many of them are stepping back from the most important shift, not just knee-jerk trimming.

The Missing Piece: Domain Intelligence + AI

It’s easy to champion “technical expertise.” Yes, you need it. You need data engineers, machine learning scientists, software architects. But technical skill alone is not enough. The real multiplier is when those technical experts are wedded to deep domain insight:

Capability	Without Domain Insight	With Domain Insight
Problem Framing	Builds models on generic KPIs or convenience data	Designs for business-critical levers, clarifies causal drivers
Feature Selection	Haphazard or brute-force	Informed selection based on domain logic
Interpretability & Trust	Black box, hard to justify to executives	Transparent, tied to business logic and decision-making
Adoption	Low — business hesitant to trust	Higher — domain experts can embed and evangelize

In short: AI without subject experts is like giving a blank canvas to someone who doesn’t know the picture you want to paint.

What “AI-First Subject Matter Experts” Look Like

So what do I mean by an “AI-first subject matter expert”? It’s a hybrid kind of leader (or team) who:

Knows the business inside out. They understand the domain, the key levers, the constraints, the processes, the culture.
Speaks tech fluently. They’re not just translating — they truly grasp data, algorithms, architecture, and limitations.
Can map domain problems to AI solutions. They identify which business problems can be transformed by AI (and which can’t).
Bridge between disciplines. They act as the glue — communicating between data science, engineering, operations, and business leadership.
Iterate fast, focused on value. They launch small, prove value, learn, and scale — always tethered to business outcomes.

These hybrids aren’t just a nice-to-have; increasingly, they are the only way ambitious AI deployments survive beyond the POC stage.

Why This Approach Counters Pullbacks — And Inspires Confidence

Let me argue why adopting this mindset is exactly what can counter the wave of pullbacks:

Risk is reduced. When domain experts shape the problem, you’re less likely to chase vanity metrics or build irrelevant models. Cancellations often happen when efforts seem speculative — this bridges that gap.
Trust builds faster. Business stakeholders see that the AI effort isn’t some “black box magic,” but grounded in domain logic they understand.
Outcomes tie to real impact. When you build with the core levers, you produce results that executives care about — revenue, margin, operation efficiency — rather than side metrics.
Scalability is more plausible. A domain-informed design is more maintainable, explainable, and aligned with core processes, improving the odds of scaling vs. being shelved.

In effect, you turn your AI efforts from things that can be cut as “nice to have” to must-have capability, deeply embedded in how value gets created.

Caveats & Practical Steps

To be fair, this vision isn’t trivial to realize. Some pitfalls:

Hybrid talent is rare. It’s hard to find people deeply fluent in both business and tech.
Organizational resistance. Many firms still structure data/AI in silos, with no direct ties to business units.
Overreliance danger. Domain experts must be careful not to override technical best practices or bias solutions. The partnership must be true collaboration.

If you want to get started, here’s a rough roadmap:

Identify high-leverage domains. Pick core value chains or levers (e.g. supply chain, pricing, customer churn).
Staff hybrid roles. Either train domain experts in AI or embed domain folks with data teams.
Run rapid pilots. Focus on small but meaningful problems, tie to real metrics.
Build shared language and trust. Workshops, cross-functional teams, shared accountability.
Scale with governance. Use modular architectures, clear measurement, explainability, monitoring.

Closing Thoughts

Yes, it’s understandable that large companies, under pressure, would pare back — especially areas that feel speculative or disconnected. But I would argue that the things they should double down on are exactly those that blend domain insight and technical strength. Because that’s where real transformative AI emerges, not as a sci-fi afterthought, but as a core engine of advantage.

In the end, big companies shouldn’t just survive the pullback — they should pivot toward the deeper, harder, higher ROI path. That’s where I believe the real leaders will emerge.

smartscale