I sat down with Aaron Levie on the O’Reilly AI Codecon two weeks in the past. Aaron cofounded Field in 2005, and 20 years later, his firm manages content material for about two-thirds of the Fortune 500. Aaron is without doubt one of the few CEOs of an incumbent enterprise software program firm considering deeply in public about what AI means for the whole enterprise stack. There are lots of people who’re constructing firms from the bottom up with AI, others who’re dragging their toes adapting present enterprises to it, after which there’s Aaron. He sits in a type of Goldilocks zone, enthusiastic however not uncritical, partaking within the exhausting work of adapting AI to the enterprise and the enterprise to AI.

The engineering demand paradox

I began out by asking about one thing from Lenny’s E-newsletter that Aaron had retweeted. Regardless of all of the doom rhetoric, TrueUp information exhibits software program engineering job postings are at a three-year excessive. Product supervisor jobs are manner up. AI jobs as a complete are manner up.

The precise information could also be extra equivocal than the TrueUp report suggests. The trustworthy learn of the literature as of spring 2026 (Brynjolfsson et al., Humlum and Vestergaard, BLS Software program Builders, BLS Laptop Programmers) is that one thing actual is going on to entry-level software program work, that it’s occurring quicker than most earlier expertise transitions, that it has completely different results relying on which job code you take a look at, and that it’s not but clear whether or not the web impact on complete software program employment can be damaging, impartial, or finally constructive. Nonetheless, the TrueUp report was a set off for the dialogue that adopted.

Aaron famous that engineers have traditionally been concentrated at tech firms as a result of the price of a software program mission was too excessive to justify anyplace else. But when brokers make an engineer two to 10 occasions extra productive, all of the software program initiatives that had been by no means economically viable abruptly turn into viable. Demand doesn’t shrink. It diffuses throughout the whole economic system. In his tweet, he referred to as it “Jevons paradox occurring in actual time.” In our dialog, he stated:

“What’s going to occur is the whole world goes to be taking a look at all of the potential software program that they construct. They usually’re going to begin to say, Oh, I can lastly justify going out and doing this kind of mission the place I couldn’t earlier than.”

Engineers empowered by AI brokers received’t simply construct software program for IT groups. The entire addressable position of the engineer expands from the expertise division to each perform within the enterprise. They’ll be wiring up automation for advertising and marketing, authorized, accounting, and each different company perform.

He’s completely proper. Go searching in any respect the crappy workflows, the crappy processes, the unimaginable overhead of issues that should be easy. You suppose firms ought to lay off their builders to cut back prices when there’s a lot shitty software program on the market? Actually? There’s a lot that must be improved. He had an incredible line: “Silicon Valley is spooked by its personal expertise.”

Over to me: The rhetoric from the labs about job destruction is actively counterproductive. I used to be speaking just lately with somebody in healthcare who described a hospital system making an attempt to fill a large gap from lowered Medicare funding. They see AI as a technique to achieve effectivity of their again workplace to allow them to unlock extra sources for affected person care. And naturally the union is combating it as a result of they’ve been advised AI is a monster that’s going to take their jobs. In case you inform a unique story, one about making the system higher and serving extra folks extra affordably, that’s one thing folks can get behind. We have now to alter the narrative.

Context, not connectivity, is the true downside

I additionally requested Aaron whether or not protocols like MCP are making context moveable sufficient to erode aggressive moats. He agreed that the business has broadly converged on openness and interoperability (with some toll cubicles to work via). However getting your techniques to speak to one another doesn’t resolve the tougher downside of getting your information structured in order that brokers can really discover the fitting info on the proper second.

“If it’s in 50 completely different techniques and it’s not organized in a manner that brokers can readily benefit from, what you’re going to be is on the mercy of how properly that agent finds precisely the context that it must do its work. And also you’re type of simply rolling the cube each time you do a workflow.”

He predicts a decade of infrastructure modernization forward, which sounds about proper. At O’Reilly, I hold operating into this myself. I’ll see a job that’s excellent for an agent and shortly uncover that the information I would like is scattered throughout 4 techniques and I’ve to leap via hoops to determine who is aware of the place the information is and how one can get entry. A pal operating a big (however comparatively new) enterprise that’s turbocharging productiveness and repair supply with brokers advised me just lately {that a} massive a part of his group’s success was doable as a result of they’d spent quite a lot of time getting their information infrastructure so as from the beginning.

IMO, quite a lot of the tales you hear about OpenClaw and different harbingers of the agent future could be deceptive in an enterprise context. They’re doing greenfield setups, largely operating client apps with well-defined interfaces, and even then, it takes weeks to arrange correctly. Now think about agentic frameworks for firms with 1000’s of workers, lots of of legacy apps, and deep wells of proprietary information. A decade of infrastructure modernization is beneficiant. With out assist, many enterprises may have issue making the transition.

Engineering the trade-offs

I introduced up Phillip Carter’s “two computer systems” framing, that we’re now programming a deterministic laptop and a probabilistic laptop on the similar time. Abilities are a bridge, as a result of they’ve each context for the LLM which may work probabilistically and instruments which can be constructed with deterministic code. Each techniques coexist and work in parallel.

Aaron referred to as the boundary between the 2 computer systems “the trillion-dollar query.” When does a course of cross the edge the place it ought to be locked into repeatable, deterministic code? When ought to it keep adaptive? Mortgage processing must work the identical manner each time. Worker HR queries could be probabilistic. And the irony, as Aaron identified, is that making these trade-offs accurately requires deep technical understanding. AI makes the sector extra technical, not much less.

I added that generally this judgment is a consumer expertise query, generally a price query. You are able to do one thing with an LLM, however it is likely to be rather a lot cheaper with canned code. At different occasions, though the LLM prices extra, the flexibleness of a liquid consumer interface is much better.

That is additionally a locus of creativity. What you carry out of AI is what you carry to it. Steve Jobs wasn’t a coder, however he knew how one can get essentially the most out of coders. He would have gone nuts with AI brokers, as a result of he was the essence of style and judgment and setting the bar.

The place startups win

I requested Aaron concerning the dangers to present enterprises from greenfield AI startups that may simply transfer quicker, reinventing what the incumbents do with an AI native answer, with out all the luggage. He replied:

“If there’s already a considerable quantity of the information for that exact workflow in an present system, and the incumbent is agile sufficient and responsive sufficient, then they’re in a very good place to construct both the options or to monetize that set of labor that’s going to be carried out….What brokers are actually good at is automating the unstructured areas of labor, the messy, collaborative human-based components of labor, the tax course of, the authorized overview course of, the audit and danger evaluation technique of your entire contracts and unstructured information. And so in these areas, there’s no incumbent. The one incumbent is probably going skilled companies corporations. In order that’s the place I’d favor startups.”

Software program startups like Harvey are already taking companies domains and constructing brokers for them. Nevertheless it’s not simply software program startups. Aaron additionally sees a lot of alternative for AI-native regulation corporations, accounting corporations, and advert businesses that may throw away legacy workflow, begin from scratch, and ship two to 5 occasions the output at decrease value may have an enormous benefit.

I did push again with some extent I feel is underappreciated: Present enterprises face an actual danger that the group will attempt to stuff AI into present workflows moderately than asking what the AI-native workflow could be. Individuals are connected to their jobs, their roles, the org chart. We have now to wrestle with that actually if we’re going to actually reinvent what we do.

People get context without spending a dime

One in every of Aaron’s factors about brokers is that people carry an unlimited quantity of ambient context that brokers lack. You recognize what constructing you’re in and who else works there and what they do. You recognize the assembly that simply occurred the place a group modified course on a technique that hasn’t been written down but. You’ve gotten 20 years of collected area information. All of that’s free context that we’ve by no means needed to formalize. As he put it, “We’ve by no means constructed our enterprise processes in a mannequin the place we assume that there’s a brand new consumer in that workflow that appeared one second in the past and in underneath 5 seconds, they should get the entire info doable to do this job.”

He urged that a technique to consider brokers is as new workers who’re specialists however arrive with zero context and should be absolutely briefed. And the context must be exact, not simply complete. Give an agent an excessive amount of context and it will get confused. Give it too little and it rolls the cube. SKILLS.md and AGENTS.md recordsdata are makes an attempt to supply precisely the surgical context an agent wants for a selected course of.

However 99% of information work doesn’t have an AGENTS.md file, he famous. The information is in every single place. The context is in every single place. So in an present enterprise, it’s a must to reengineer workflows from the bottom as much as ship the fitting info to brokers on the proper second.

Aaron summed up Field’s strategic pivot in a single sentence: swap the phrase “content material” for “context” and the remainder of the technique stays the identical. Enterprise context lives in contracts, analysis supplies, monetary paperwork. That’s all enterprise content material however it isn’t all the time simply obtainable as context. The evolution is making brokers first-class residents alongside folks as customers of that content material. This very a lot maps to what we’re excited about at O’Reilly too.