Trust Is Not a Digital Asset — And It's the One Thing Your AI Strategy Is Burning Through

What Companies Are Actually Cutting

Ask a CEO what their AI strategy is cutting and you'll get the sanitized answer: repetitive tasks, manual data entry, first-draft work. The stuff nobody wanted to do anyway.

That's not what's actually happening on the ground. I've reviewed restructuring plans at three companies this year where the euphemism was 'efficiency gains through automation,' and the reality was eliminating entire cohorts of analysts, coordinators, and associates who used to sit in client meetings, vendor calls, and cross-functional syncs.

Those roles looked like overhead on a spreadsheet. They were never overhead. They were the pipeline.

Fast Company ran a piece recently arguing that 80% of knowledge work is automatable, but the remaining 20% (judgment, relationships, domain authority) is irreplaceable. I don't disagree with the math. I disagree with how companies are responding to it. They're optimizing the 80% and quietly defunding the 20%.

Trust as an Enterprise Moat

At Novartis, I managed relationships across 90 countries and 1,200-plus websites. The technical infrastructure mattered. It was not what kept those markets functional when something broke at 2am local time.

What mattered was that a country lead in Brazil trusted our team enough to call before escalating to legal. What mattered was that a regulatory affairs contact in Japan would give us a heads-up on a policy shift before it hit the wire. None of that trust was in a system. It lived in people, built over years of showing up, following through, and being right often enough to be believed the next time.

Trust is the only moat left in most industries. Your pricing gets matched. Your features get copied within a quarter. Your tech stack is replicable by any competitor with a budget. The relationships your people have spent a decade building, the domain knowledge that lets someone read a client's hesitation and know exactly what's driving it, that's not replicable. It's also not on your balance sheet, which is exactly why it's the first thing finance teams overlook when they model AI-driven savings.

The Relationship Pipeline You're Eliminating

Here's the mechanism nobody's pricing into the automation math. Junior people don't just produce first drafts. They sit in on the client call to write it up. They shadow the vendor negotiation. They get looped into the internal debate about strategy, badly, at first, and gradually get better at reading the room.

That's not inefficiency. That's the entire mechanism by which institutional trust and domain authority get created in the first place. Senior judgment doesn't materialize. It's manufactured, slowly, through thousands of small reps most companies now consider too expensive to fund.

I saw this at Razorfish while building our consulting division. The junior strategists who sat through a hundred mediocre client meetings became the senior partners who could walk into a room and know, within ten minutes, what the client actually needed versus what they said they wanted. You cannot shortcut that. You can only fund it or defund it.

When you automate the first draft and eliminate the seat at the table that used to come with producing it, you're not just cutting a cost line. You're severing the pipeline that turns twenty-five-year-olds into the trusted advisors your clients call directly in year twelve.

The Long-Term Value You're Destroying

This shows up in the numbers eventually, just not on the quarter you're optimizing for. I've watched companies post strong margin improvements for two, sometimes three years after aggressive automation of junior roles, and then hit a wall: no bench. No one ready to step into senior relationship roles because the pipeline that built judgment got cut five years earlier.

By the time it shows up in client churn or in a failed account transition, the executives who made the cut are two roles removed from the consequence. Nobody connects the dots because the lag is too long and the attribution is too diffuse.

A pharma client I worked with cut junior medical science liaison headcount by 40% in favor of AI-assisted content generation for field teams. The content got faster and cheaper. Three years later, they couldn't explain why their KOL relationships had gone cold, why physicians who used to take their calls first were now taking a competitor's. The answer was simple: nobody had been in the room building those relationships for three years. The AI wrote good content. It never once built trust with a human being.

That's the pattern. Automation shows up as savings immediately. Relationship erosion shows up as revenue loss two or three years later, disguised as market dynamics or competitive pressure, when it was actually a staffing decision made in a budget cycle nobody remembers.

What to Protect, and How

None of this is an argument against automation. Automate the 80%. You'd be negligent not to. The argument is that the 20% needs deliberate, funded protection, not leftover budget.

A few things I'd put in front of any CEO or CDO building an AI strategy right now:

First, model the relationship pipeline explicitly. If your junior roles are being automated, name the specific mechanism by which the next generation of senior relationship-holders will get built. If there isn't one, you have a gap, not a strategy.

Second, treat trust-building activities as protected headcount, not discretionary. Client-facing shadowing, cross-functional exposure, the unglamorous meetings that used to be junior busywork. Budget them the way you'd budget R&D: as an investment with a payback period longer than a fiscal year.

Third, measure relationship depth the way you measure any other asset. Who on your team has direct, trusted relationships with which client or partner contacts? What happens to that relationship if that person leaves tomorrow? If you don't know the answer, you're carrying risk you haven't priced.

Fourth, ask a harder question of every automation decision: not 'what does this save,' but 'what does this cost us in five years that we can't currently see.' The companies that get AI strategy right treat the 20% as the thing they're protecting the ability to automate the 80% in service of, not the thing left over after automation eats the budget.