A few years ago, most board and executive conversations I had about AI were tactical.

“Can we automate this process?”
“Where’s the ROI?”
“Can we shave some cost?”

Today, the questions are blunt:

“What happens to our business model if we don’t move on AI?”
“What happens when a competitor rebuilds our value chain with agents end-to-end?”
“How fast could we pivot if we had to?”

That shift matters. AI has gone from side experiment to default expectation. Your customers assume you’re using it. Your competitors are exploring it. At that point, AI stops being a technology discussion and becomes a strategic one.

This isn’t about sprinkling AI on old workflows. It’s about what business you’re really in and how long your current model can survive.

From efficiency project to “what business are we in?”

When we started what is now Performance AI back in 2021, most clients wanted a more efficient version of what they already did: faster back-office processes, smarter document processing, more timely reporting.

Valuable, but incremental. The assumption was: “Our business model is fine; we just want it to run better.”

What has changed over the last couple of years is agency.

AI is no longer just analyzing. It’s acting. Agents can plan, take actions across systems, coordinate with other agents, and only escalate to humans when they need help. When software can read a contract, update multiple systems, notify the right people, and catch edge cases, you’re not tweaking the old design – you’re redesigning how the business works.

That forces a harder question:
If we were building this business from scratch today, in a world where agents exist, would we design it the same way?

In many industries, the honest answer is “no.” That’s where disruption starts.

How AI is reshaping the value chain

When I work with leadership teams, I break the impact of AI into three layers:

1. Customer interface – how value is experienced

2. Decision and coordination – how value is orchestrated

3. Execution and operations – how value is delivered

AI is changing all three at once.

1. Customer interface: from channels to conversations

We’re moving from static channels to ongoing, context-aware conversations.

In healthcare, an agent can coordinate appointments, insurance, prep instructions, and follow-up, and bring a human in only when something looks unusual. In financial services, an AI “co-advisor” can explain what changed in a portfolio, why it matters, and what to consider next.

Whoever owns that persistent, intelligent interface starts to own the relationship.

If you let a third-party AI layer sit between you and your customer, you risk becoming the infrastructure behind someone else’s experience and margins. You still do the work; someone else controls the touchpoint, the data, and eventually the pricing power.

2. Decision and coordination: the new operating system

For decades, the “brain” of the business lived in managers, spreadsheets, email threads, and a few rigid rule engines.

Now that brain is shifting into agentic orchestration.

Agents triage demand, reconcile data, propose resolutions, and coordinate both humans and other agents. They don’t just generate dashboards; they move work through the system.

At Performance AI, this is our focus. Our modular agent platform plugs into existing systems — CRM, ERP, EMR, data warehouses — and acts as an AI backbone across workflows. We’re not trying to replace everything you have; we’re trying to connect it and make it behave like one intelligent system.

Competitive advantage is starting to show up here:

• How fast can you re-route work as conditions change?

• How dynamically can you price and allocate?

• How well do you blend human judgment with machine action?

3. Execution and operations: from FTEs to digital workforce

At the execution layer, the question is shifting from:

“How many people do we need?”

to:

“What combination of humans and agents do we need?”

Agents now:

• Process documents from ingestion to action

• Monitor compliance and surface only what needs review

• Run forecasting and scenarios on demand

Once a company gets a few agents working well in one domain — finance, revenue ops, logistics, compliance — they realize these are building blocks. You don’t want a different one-off bot per team. You want a stack of agents you can reuse and scale.

At that stage, the question stops being:

“Can we save 10–20%?”

and becomes:

“What happens when a competitor runs this same business with a fundamentally different cost curve and cycle time?”

New business models AI makes possible

Efficiency is just the starting line. The real story is the business models AI unlocks. I see four patterns repeatedly.

1. Outcome-based services

If agents can instrument workflows and tie activity to outcomes, you can move beyond seat licenses and vague platform fees.

You can price on shared savings, resolutions, or success metrics. If you trust your AI backbone, you’re more willing to get paid when the customer wins.

That’s attractive to buyers and harder for low-value competitors to copy.

2. AI-as-backbone platforms

Every industry will have a few players who become the AI backbone: the agent layer tying together core systems, data, and customer touchpoints.

That’s the role we’re building toward with Performance AI: a modular, compliance-ready backbone that coordinates data, workflows, and teams. When you own that backbone, you’re not just another tool. You become the coordination layer of the ecosystem.

That’s a very different position from “yet another SaaS product.”

3. Data network effects

Agents embedded in real workflows generate rich signals:

• How work actually happens

• What “good” looks like

• Where exceptions occur

• Which actions lead to better outcomes

Handled well, that enables domain-specific models and products: pre-tuned agents, best-practice workflows, aggregated and anonymized benchmarks.

Handled badly, it becomes a privacy and trust issue. Handled well, it becomes a compounding data advantage with every deployment.

4. Human+AI bundles

Some of the strongest AI offers I see are not pure software. They bundle:

• Agents

• Domain experts

• Implementation and change-management

So the outcome is a redesigned way of working, not just a new interface.

The customer doesn’t say, “We bought an AI tool.” They say, “We changed how this part of the business works, with support.” That feels more like a managed service or co-pilot subscription than classic SaaS — and the economics follow.

How I feel about where this is going

I’ve lived through a few big technology waves. When I worked on the Android SDK, we watched mobile reshape industries over a decade. With AI — especially agentic AI — that timeline feels compressed.

What used to require multi-year system overhauls can now be piloted in weeks by layering agents on top of existing systems and scaling what works. That’s exciting and unforgiving at the same time.

Personally, I feel three things:

• Excitement, because we finally have the tools to make the “smart enterprise” real.

• Urgency, because there is still a window where AI is a differentiator, not just table stakes — and that window is closing.

• Responsibility, because we’re reshaping how work is done, who does it, and who benefits. If we get this wrong, we damage trust, not just margins.

My advice, as a CEO building in this space every day, is straightforward:

• Don’t treat AI as a side project.

• Don’t be blinded by demos and ignore your operating model.

• Don’t wait for perfect data or perfect org readiness. Pick one critical workflow, deploy agents against it with clear metrics, learn fast, then scale.

Treat AI as a design constraint for your next business model, not as a bolt-on feature.

Because a few years from now, your customers, competitors, and investors won’t ask,
“Did you adopt AI?”

They’ll ask,
“Why does your business still look like it did before AI?”

You want a very good answer to that question.