Something for the weekend - 'Who does number 2 work for?!' How agents will challenge enterprise boundaries

It feels to me — should the greater good have any say in the agentic debate — that the natural endpoint of agentic technology will be the creation of deeply personalized systems that understand and augment the individual strengths and weaknesses of each and every one of us. At least that’s the vibe I get from the luminaries of the industry.

But if that's our goal — for agents to maximize human potential — then surely it also stands to reason that these agents must belong to us as individuals, not to the big corporations we shop from or work for. They must be deeply personal, knowledgeable, have shared values, and — most of all — respect our absolute right to privacy.

Having seen what social media has done to pervert the incentives of online discourse, I'm pretty sure nobody wants those same big-tech algorithms twiddling our innermost thoughts for outrage and engagement stats. It feels logical therefore that these agents should belong entirely to us — since they won’t only support us but also become our primary window onto the online world.

The real promise of agents isn’t to enable you to shop from an armchair without so much as raising a fat finger for 'one-click' purchases. Instead, the real promise is to empower us by reducing the co-ordination costs of anything we want to get done — by handling research, context evaluation, curation, and action on our behalf. Effectively amplifying our capacity and transforming our capability to pursue our individual goals by getting to know us and acting in our individual best interests.

Facilitating this requires a reorganization of the online world — making it convenient for agentic knowledge discovery, actions, and oversight. As we move into the agentic era, who wants to replace context switching between applications with context switching between agents? To replace the existential dread of a 100+ active chrome tabs with the cognitive overload of 100+ active agents?

How much simpler it will be to have a single agent that co-ordinates all of that for us, which the Web interacts with as our online avatar.

What's the end game?

But if that’s the case — and I'm getting to my point here — who will want to go to work and stop using their super-supportive personal agent in favor of a fragmented landscape of enterprise agents and co-pilots that don't know their personal preferences? Their strengths and weaknesses? Their communication and information processing styles?

In other words — if you want to employ me, employ my agent. Because otherwise you're employing only a fraction of my capability.

For organizations, this shift isn’t just a philosophical thought experiment — it’s the logical intellectual endgame of agentic logic.

As employees arrive with personal agents that know their working style, memory, and decision habits, enterprises will have to decide whether to wall them off or let them participate.

This means rethinking data boundaries, compliance frameworks, and even employment contracts. Identity, access, and audit trails will no longer be tied to a person alone but to their extended cognition — the agent that works alongside them, has equal access to institutional knowledge, and may walk out the door when they do.

In short, agentic participation won’t just be a spicy AI tweak to the firewalled and single-purpose transactional enterprise applications people use today — it will redefine what it means to employ someone at all.

Which suggests to me that the true long-term mindset shift organizations need to make — if they genuinely believe that agents are going to be 'a thing' — is not how to introduce loads of agents but how to reconfigure their operating models and data foundations to enable agentic participation — whether those agents are outside or inside the boundaries of the company.

The third act

This may sound revolutionary, but we’ve been here before. Because every technological shift that reduces co-ordination costs follows the same three-act play.

First companies try to co-opt the shift and apply the new technologies to improve the efficiency of their old model, yielding marginal gains. "I know — let's put containers on our existing sailing ships as that'll be more efficient!"

Secondly in a panic they switch to distribution — opening up to new ways of integrating with customers and partners to protect market participation, which might win them some time but is kind of lipstick on a pig. "I know — let's put our monolithic legacy system online as then it will be cloud-native!"

And finally those destined to dominate the new economic environment arrive at the third and final act — re-composition. They start from first principles and reconfigure their operating models for the new possibilities of radically lowered co-ordination costs.

Today we're largely in the first phase of the agentic wave — inappropriate attempts to implement 1:1 replacement of people with agents, without changing the fundamental operating model.

There's an increasing (but early) discussion about how agents will consolidate the experience layer and require companies to optimize distribution and transactions for agents, with agentic browsers an early shot across the bows. What, essentially, will it mean when people transact directly within AI models rather than company-controlled websites or apps?

And then there's the actual fundamental reordering of the economy around this shift. The realization that the externally owned agents interacting with an enterprise will need access to its data, knowledge, and actions. Even when working within the enterprise, they will be working for its employees, partners or customers.

The macro effect of adding this new layer of intelligence is the removal of co-ordination costs for those augmented by the technology — and the compounding benefits they get from increasingly context-rich agents that know them, their history, and their preferences. Agents that are increasingly an inseparable part of their ability and sense of self, and that will be coming with them to work inside your company whether you're ready or not.

But as the fictional spy Austin Powers once shouted to a man with his head in a toilet — "Who does Number Two work for?!" 

When it comes to agents the deeply profound and unsettling answer probably won't be who you thought it was.