Your best people were always better than you knew

For thirty years the move has been to outsource.

When you needed capability you didn’t have inside the building, you hired someone outside the building. The agency did the campaign. The consultancy did the strategy. The firm did the analysis. The training vendor did the training. The reasoning was always the same, and in the companies I worked with, it was usually said out loud in roughly the same words: our team can’t do this work.

Take that seriously. It was true.

It was also true about the wrong thing. The gap wasn’t where you thought it was. And the gap you actually had just closed.

Start further back than 2026. The insource-outsource question isn’t an AI question. It’s one of the oldest questions a firm can ask about itself.

In 1927, Henry Ford opened the River Rouge complex near Detroit and built what is still, almost a century later, the canonical image of vertical integration. Iron ore came in on one end. Finished automobiles rolled out the other. Coke ovens, a glass plant, a steel mill, a power station, a rail yard, ninety-three buildings on the same site. Ford believed the way to control quality, cost, and timing was to own the whole chain. For about forty years most large American industrials thought the same way. Bring it inside. Keep it inside.

Ten years after Ford opened Rouge, Ronald Coase published a short paper called The Nature of the Firm. He asked a question everyone had been walking past: if markets are so efficient, why are there firms at all? Why not just contract everything? His answer, which eventually won him a Nobel, was that every transaction carries hidden friction — finding a counterparty, negotiating terms, enforcing the agreement — and when that friction is high enough, it’s cheaper to bundle people inside a firm and pay them salaries than to keep renegotiating with the market. The firm exists to dodge transaction costs. Make-vs-buy was now a proper theoretical question, and the answer depended on which friction was cheaper at the moment you asked.

For another forty years, the answer mostly stayed “make.”

Then the answer flipped. In 1981, Jack Welch took over General Electric and said in plain language what a generation of CEOs had been thinking in private. Every business unit had to be number one or number two in its market, or GE would fix it, sell it, or close it. Anything that wasn’t core got externalized. Welch didn’t invent the outsourcing era, but he gave it the moral authority of a dominant firm’s CEO saying it into a microphone. Within a decade, the language of “core competency” had swept the Fortune 500.

Nike had been running ahead of that curve since the 1970s, owning the design and the brand and letting contract factories in Korea, Taiwan, and later Vietnam make the shoe. By the 1990s and 2000s, McKinsey and Accenture and Deloitte had grown into the largest purchasers of ambitious graduates in the country, because every one of their client firms had stopped trying to do its own thinking in-house. Strategy, implementation, transformation — all outside the building, paid for by the hour, delivered in bound decks.

That’s the arc. Thirty years of offshoring, outsourcing, disaggregating, and paying someone else to do the thing your team used to do. And it worked, by the metric the era used to judge itself. Margins went up. Balance sheets got lighter. Headcount stayed flat while revenue grew.

Then it started to flip again.

Apple announced the M1 chip in November 2020 and did something conventional wisdom had called impossible for twenty years: it designed its own silicon, in-house, and won. Intel had been the default for a generation. Apple walked away and built something better by bringing the work back inside. Two years later the CHIPS and Science Act committed $52 billion to semiconductor manufacturing in the United States. The pharma industry started talking seriously about reshoring active ingredients after a decade of quiet dependence on a handful of plants in India and China. “Resilience” replaced “efficiency” as the word executives used when they wanted to sound serious.

The pendulum had started to swing back toward in-house even before generative AI showed up. AI just accelerated the physics.

The thirty-year outsourcing era got one thing right and one thing wrong.

It got the production question right. In 1995, the marketing director at a mid-sized company genuinely could not ship the volume and quality of work that a good agency could ship. Not because she wasn’t smart. Because producing that work required specialists she didn’t have: designers, copywriters, media planners, traffic managers, a whole scaffolding of production capacity. She didn’t have any of it. Neither did her team. The agency did. Paying the agency was the rational call.

Same logic in finance, training, strategy, research, analytics. A real gap existed between what the internal team could make and what an outside firm could make. The outside firm had the tools, the workflow, the specialists, the production capacity. The internal team had a laptop and a quarterly goal.

What the era got partly wrong was the shape of the gap.

Part of it was real. Agencies had craft your team didn’t have — working knowledge of what makes a campaign actually work, a brand system actually cohere, a strategic analysis actually land. The specialist firms had spent thirty years accumulating that craft, and most internal teams had not. Craft was a genuine piece of what got bought, and pretending otherwise would be dishonest.

But craft wasn’t the whole gap, and in most engagements it wasn’t even the biggest part. The internal team almost always understood the customer better than the outside firm ever would, because they were in the building, on the calls, watching the numbers, absorbing three years of context the agency would pick up across the first six weeks and then charge to re-learn every time the account manager turned over.

What got sold as one bundled thing — we’ll handle it — was actually three layers stacked together. Production capacity. Craft expertise. And, underneath both, a quiet assumption that the thinking had to come from outside too.

The production layer was the biggest part of the bill. The craft layer was real. The thinking layer was the lie.

The marketing director didn’t lack the strategic instinct. She lacked the designer. Her finance lead didn’t lack the model. He lacked the analyst who could build it. The product manager didn’t lack the customer insight. She lacked the person who could turn it into a competitive brief. In every case, what was missing was the output layer, not the thinking layer.

The thinking was inside the building the whole time. The output layer was outside the building. That was the bargain, and for thirty years it was the honest bargain. You paid the firm for what you didn’t have the capacity to produce. They didn’t out-think your team. They out-produced your team.

Because production capacity was the binding constraint, nobody could tell the difference. The firm’s output was better than the team’s output, so it looked like the firm was better than the team. A whole generation of marketing directors and finance leads and ops leaders got trained to believe — quietly, in the part of the brain that holds professional self-image — that the outside firm was more capable than they were. That belief got reinforced every year the arrangement persisted. Eventually it became part of how the organization understood itself.

It was wrong. It was a confusion of production with capability. And the tool that was doing all the confusing just stopped being scarce.

The production gap has collapsed.

Not “is collapsing.” Has collapsed, for a broad enough band of knowledge work that a sensible leadership team should already be assuming it for planning purposes. An internal marketing director with solid instincts and access to Claude, Canva’s current stack, and a good AI-literate operator inside her team can ship at a level that would have cost her six figures and ninety days to get from an agency in 2022. A finance lead with an LLM can build the model he used to have to brief. A product manager can draft her own competitive analysis in an afternoon.

The output layer — the layer that used to be outside the building — is now inside the building, on a subscription the company already pays for.

What’s revealed underneath, as the production scaffolding comes down, is everything that was never the gap in the first place. The judgment. The care. The context. The three years of watching the numbers. The product manager’s instinct about which feature is going to bomb, which she has never had the production capacity to write up but has been right about eleven times in a row. The operations lead’s sense of which process is quietly eating the company from the inside. None of this was ever outsourceable. The agency couldn’t buy it. The consultancy couldn’t replicate it. It was always in the building. The production bottleneck just hid it.

Your best people were always better than the arrangement made them look. Not every firm has a roster like that — some teams genuinely lack the judgment and context the agency was papering over, and for those teams this moment is going to be painful. But enough firms have those people, underused for years by a system that couldn’t afford to use them fully because the production tax was prohibitive. The production tax just dropped to roughly zero. Enough of your competitors are about to find their people for the rest of the market to feel the pressure.

That’s the opportunity. A firm that sees the reveal clearly can redeploy a layer of human capability it has been paying for, and underusing, for its entire existence. The specialist middle — the agencies, the consultancies, the training vendors, the advisory shops — thins. The internal-generalist layer deepens. Work that used to require three vendor contracts and a six-month timeline now happens between Tuesday and Thursday of the same week, done by the people who know the company best, because the people who know the company best can finally produce.

This isn’t “AI replaces your people.” It’s the opposite story. AI reveals your people. Reveals who on your team was actually good, who has been carrying more than the job description showed, and who — honestly — hasn’t. Some of what gets surfaced is going to be uncomfortable. That’s the work.

Getting there takes three things, and the investment has to land on all three.

The first is the tools. AI fluency. Knowing which model for which task, how to prompt it, how to chain it into a workflow, where it breaks and where it shines. This is the layer everyone talks about because it’s the visible layer. It’s also the commoditizing layer — Anthropic and Canva and Google are teaching it to the whole market at the same time, inside the products they sell.

The second is the craft expertise. What actually makes a good marketing campaign, a clean brand system, a training program that changes behavior, a strategic analysis worth acting on. Craft is what the consultancies and agencies sold for thirty years. It didn’t go anywhere. The AI era didn’t delete it. AI just separated the craft from the production, which is what made it look for a minute like the craft itself was being automated. It wasn’t. The craft still has to come from somewhere.

The third is the set of human capacities the AI era actually demands. Judgment. Discernment. The taste to know when the model is producing something polished and wrong. The initiative to ship something novel instead of the thing the model defaulted to. The creativity to see an angle the training data hasn’t seen yet. The ability to read a situation and pick the right move when the model wants to hand you the average one. These are the capacities the industrial era systematically suppressed because it didn’t need them at scale. It does now.

Invest in all three and you get an internal team that produces at agency standard without the agency, on a problem the agency didn’t understand as well as you did anyway. Invest in only one leg and you get a specific, predictable failure.

You’ll see all three failures in the market over the next eighteen months.

Invest only in the tools and your team is going to be doing dentist flyers in Canva. The AI subscription doesn’t know what good marketing looks like, and neither does anyone using it without the craft underneath. You’ll get fast output that looks like every other company that bought the same subscription. Volume without taste. Confidence without judgment. The machines will happily produce a brochure that would embarrass a mid-tier dental practice, and your team won’t know why it’s wrong, because nobody taught them the thing that makes it wrong.

Hire only the outside expertise and you hit the dilemma. If your vendor isn’t using AI, they’re overcharging you for work that’s gotten cheaper. If your vendor IS using AI, they’re charging you for work your team could do with the same tools. Either way, something is wrong.

That was the bargain when your team couldn’t produce. It isn’t the bargain now. The bundle the vendor used to sell was real — a production line, a room of specialists, a deliverable your team couldn’t make. That bundle is coming apart. The AI half moved to a subscription you already pay for. The craft half is still worth paying for, but as transfer, not as retainer. You’re paying for the tool layer twice — once inside your subscription, once inside theirs.

The firms that survive this decade will sell craft transfer, not craft delivery, and the firms buying craft delivery are going to look, in hindsight, the way firms that kept buying mainframes through the 2000s look now.

Hire only a coach and you’ll get a team that’s gotten more reflective but no more capable. The coach doesn’t know anything about marketing or artificial intelligence, and being reflective about work you can’t execute is its own kind of trap. The human-capacities layer is real and it’s important, but by itself it’s a therapy practice, not a capability build. You’ll have a team that journals about its growth while the work goes unshipped.

The triple works because each leg covers what the others can’t. Tools without craft produces garbage at speed. Craft without tools produces elegance no one can ship. Either of those without the human-capacities layer produces a team that can technically do the work but won’t exercise the judgment the work actually requires. Tools plus craft plus human capacities is the smallest complete set. Miss any one and the investment underperforms in a specific, nameable way.

I wrote a book, The Work of Being. The argument was that the twentieth-century workplace had systematically suppressed the most human capacities of the people inside it — judgment, creativity, initiative, discernment, the willingness to think independently and act on the thought — because the industrial and post-industrial economies could run on compliance and predictability, and compliance was cheaper to manage. The system rewarded the parts of people that behaved like machines and punished, or quietly underused, the parts that didn’t.

That argument was intended philosophically. It turns out to have been commercially predictive. The capacities the industrial era suppressed are precisely the capacities the AI era demands. A workforce trained over three generations to defer, comply, and produce to spec is, right now, holding tools that reward the exact opposite orientation. The people who can actually take this moment are the people who still have a working relationship with their own judgment. That’s rarer than it should be. Rebuilding it is a real piece of work, and it’s the third leg of the triple.

So: the insourcing moment is real, it’s happening, and taking it requires all three legs at once.

Your internal team has the context. They’ve been in the building for years, watching the numbers, absorbing the customer, developing the judgment the arrangement never asked them to use. The production gap that justified thirty years of outsourcing has collapsed to roughly nothing. The tools to produce at agency standard are running on the subscription the company already pays for. What’s been holding the team back isn’t capability. It’s capacity, and capacity just became cheap.

The opportunity is to stop buying production from outside and start using the production that’s suddenly free, operated by the people who always knew the company best. That takes the tools, the craft, and the human capacities the last century suppressed. Any one of them alone gets you one of the failure modes above. All three together gets you an internal team performing at a level it has never been allowed to perform at before.

Your best people were always better than the arrangement let you see. The firms that figure out who those people are, equip them with the tools and the craft, and activate the human capacities the last century suppressed — those firms are about to make your agency contracts look like a cost center. The firms that don’t are going to spend the next decade paying for craft delivery on a layer the market no longer requires.

The reveal is happening either way. The only choice is whether it happens inside your firm or to it.

— Paul Welty