The forecast this session leans on is worth examining before accepting: a projection that over 33 per cent of enterprise software will include agentic AI by 2028, up from around 1 per cent today (3:12).
Read carefully, that is a strange claim. It is not a prediction that a third of software will be agentic. It is a prediction about how many products will contain some agentic capability — which is a much lower bar, and one most vendors will clear by adding a feature rather than by changing anything.
The reason to be sceptical is the base. Going from one per cent to thirty-three requires either extraordinary genuine adoption or a definition loose enough that a chat assistant qualifies. History suggests the second.
The decision that actually matters
The useful framing arrives immediately after: customers face a build-versus-buy decision (4:01).
That is the real question, and it is more interesting for agents than for conventional software, because the usual reasoning does not transfer cleanly.
Ordinarily you buy commodity capability and build what differentiates you. With agents the boundary is harder to locate, because the value sits in the connection between generic reasoning and your specific context — your data, your processes, your exceptions. A purchased agent brings capability and no context. A built one starts with the context and needs the capability assembled around it.
Which suggests the decision hinges on where the difficulty is. If the hard part is the reasoning, buy. If the hard part is knowing which of your seven customer identifiers is authoritative, no vendor can sell you that.
Why agents specifically
Their explanation of what makes agents different is compact and correct: they think iteratively, create a plan, iterate and call tools (2:12).
That distinguishes an agent from a model in the way that matters commercially. A model produces output from input. An agent pursues an outcome, which means it can be sold against a job rather than against a capability — and a job is something a buyer can evaluate.
It is also why the distribution question exists at all. Nobody needed a marketplace for models; you call an API. Something that performs a job has integration requirements, permissions, and a boundary of what it will do unsupervised, and those need somewhere to be negotiated.
The part that will decide adoption
The mechanics get less attention than the forecast and matter more: access keys from the seller, with the buyer redirected at sign-up (14:04).
That is the join between purchased capability and the buyer's environment, and it is where these arrangements will succeed or fail. An agent that must be given credentials to your systems is not a purchase in the way software has been. It is closer to granting access to a contractor, and the governance of that — what it may reach, under whose authority, revocable how — is not solved by a distribution channel.
The forecast may or may not hold. The unresolved question is not adoption but what an organisation is actually agreeing to when it buys one of these.
关键数据
- 1% → 33%
- projected share of enterprise software including agentic AI by 2028 3:12
演讲章节
关键要点
- 01
The projection is that over a third of enterprise software will include agentic AI by 2028, up from around one per cent — a claim about features rather than about architecture. 3:12
- 02
What distinguishes agents commercially is that they pursue an outcome, which means they can be sold against a job rather than a capability. 2:12
- 03
The decision customers actually face is build versus buy, where the usual reasoning transfers badly because value sits in the join between reasoning and context. 4:01
- 04
The mechanics matter more than the forecast: access keys from the seller and a buyer redirected at sign-up is the join between purchased capability and your environment. 14:04
提及的实体
相关演讲
The practical counterpart to the argument made elsewhere this season that specification is what contains model entropy. Raval and Harris name the failure they are addressing precisely — a prompt-and-pray loop in which working code arrives with no record of what the model assumed, which requirements were fuzzy, what design was chosen or why, leaving nothing to review and nothing to iterate against when a defect surfaces months later. Their answer is three committed markdown artefacts: requirements written in a structured requirements syntax with acceptance criteria attached to each user story, a design document carrying technical decisions together with the reasoning behind them, and a task list whose entries cite the requirement numbers they satisfy. The traceability is the point — a reviewer questioning a decision in a pull request can follow it back through the task to the design to the requirement, all in the same repository. Notably they keep the human between each phase rather than after it, with the agent surfacing ambiguity as questions before proceeding.
Drawn from a year of engagements with more than a hundred companies, this is the most direct challenge in the season's programme to the assumption that faster code generation produces faster delivery. Mishra and Raja open with external evidence rather than their own: an industry study putting realised velocity gains in the ten to fifteen per cent range, and a controlled experiment in which developers using AI estimated themselves roughly a fifth more productive while measurement showed them a fifth slower. Their diagnosis is that both prevailing working styles fail for opposite reasons. Handing an ambiguous problem to an agent and awaiting a finished result produces a volume of code the developer must nonetheless sign for and cannot confidently review, so it stalls before production. The senior engineer's alternative — decomposing the work personally and inserting AI into narrow slots — keeps the intellectual load exactly where it was, and leaves the surrounding process untouched, so hours saved in editing are consumed by the meetings that process still requires.
Brooker builds the definition from the bottom up rather than asserting it, using a deliberately absurd arithmetic task to separate three categories: what a model computes reliably as a fixed function of its input, what merely needs to arrive in the system prompt, and what genuinely requires reaching into the world. Only the third category justifies a tool, and the distinction matters because most production disappointment comes from tools built for the first two. His working definition follows — a system given a goal that loops between inference and tool calls until it reaches one — with the observation that modern agents increasingly embed code in their definitions, not for expressiveness but because replacing inference steps with deterministic code improves reliability while lowering both latency and cost. The remainder covers what production actually demands around that loop: somewhere to run, memory that persists preferences, a gateway to internal and external tools, evaluation, and formal methods applied to policy.
The most concrete attempt this conference season to answer a question the agentic coding sessions mostly leave open: if commit counts and hours saved are the wrong measures, what replaces them? Otto's account is unusually specific about why the obvious alternative fails — summing the small time savings a platform team delivers produces figures exceeding a hundred per cent of a developer's time, and a minute returned is not code in production. Their replacement borrows from Amazon's retail supply chain, where cost to serve measures what it takes to place a package on a doorstep, and applies the same shape to software: total cost divided by units of delivery, with the unit chosen to fit the team. The supporting research is the more quotable finding — across tens of thousands of developers over five years, individual velocity reverts to the team's mean, making team velocity the strongest predictor of both individual output and perceived productivity, which is the empirical case against measuring individuals at all.
The framing statistic is organisational rather than technical: around eighty per cent of organisations expected to have platform engineering teams going into 2026, up from about forty-five per cent a couple of years earlier. The interesting part is the doubling. The problem described is teams solving the same problems separately, producing inconsistency and redundancy — dangerous not because of duplicated effort but because each independent solution has its own security properties, and the organisation's real posture is the weakest rather than the average. The most valuable content is that two financial services organisations went in diametrically opposite directions on workload identity and both are described as working, which implies the choice is determined by context rather than by a general answer. The honest note follows immediately: even with standardised patterns the result remains fragmented.
The line that explains this session comes from the customer in the last ten minutes: they are preparing for a world where metadata is how agent-based systems find the data they need and access it through the controls being built. That relocates a function — governance has spent two decades as compliance activity describing data that people locate by other means, and if agents navigate by the catalogue then the catalogue stops describing the access path and becomes it. An incomplete catalogue is a documentation problem when humans can ask a colleague; an agent has no such workaround. The most honest moment addresses the perennial failure that rules get written and ignored, with enforcement rather than publication as the argument. Generated descriptions and greyed-out classification suggestions divide the labour correctly, keeping a person accountable while removing the burden of finding candidates.