Give a factory a production line, a warehouse and a control room and you have a plant. You still do not have output. Factories run because people stand at the stations — people who know the process, spot the exception and decide what happens next.
DocReveal has had the plant for a while: one line that understands every document, a permanent memory of what was learned, and central control over all of it. Now it has the workforce. Agent Pool lets your own people define digital colleagues — in a browser, without code — and put them to work on the factory floor.
Automation stalls where the knowledge is
Most automation programmes do not fail on technology. They fail on distance.
The person who knows that one supplier always sends invoices inside a ZIP, or that a particular contract has to be flagged ninety days before renewal, is rarely the person who can build anything. Their knowledge has to be written down, handed to a queue, translated into a specification and scheduled into a sprint — and by the time it ships, the process has moved on. Every organisation carries a backlog of small, obvious, valuable automations that nobody will ever get to.
Agent Pool removes the translation step. The person who owns the process describes the job, and the factory does it.
What a digital colleague actually is
An agent in DocReveal is defined the way you would brief a new hire, not the way you would write software:
- A job description — what it is responsible for, in plain language, with your rules and your exceptions. It is a versioned document: edit it, compare versions, roll one back.
- Access — which mailbox, which folder, which system, granted from your Connection Hub. Credentials stay with the platform; the agent never sees them.
- Permissions — the specific things it may do, and which of them need a person to sign off first.
- A budget — how much work it may do, and what happens when it reaches the ceiling.
- Somewhere to ask — when it is not sure, it stops and asks a human instead of guessing.
You try it out on real cases in a playground before it touches anything, and you publish it once it behaves.
They work with everything the factory already has
This is the part that matters strategically. Agent Pool is not a separate product bolted on the side — it is the layer that finally puts the rest of DocReveal to work.
- The document truth — agents reason over validated, canonical data, and every claim they make comes back with a citation to the exact file and page.
- Your connections — the same catalogue the Integration Manager uses, with more than 1,500 ready-made actions across business applications, becomes the toolkit agents act with.
- Search and knowledge — everything the RAG layer and the graph of your entities and relationships know is available to them as a matter of course.
- Analytics and dashboards — an agent can check a governed number before it decides, instead of estimating.
Nothing new had to be connected for any of this. The capabilities were already in the factory. Agents are simply the first workers with hands.
The first one takes minutes, not a project
You start from a blueprint rather than a blank page: a first agent is chosen, connected, tested on your own recent cases and published in a single sitting.
The most common first hire is the mailroom agent. It watches a shared mailbox, decides which messages actually contain documents worth keeping — and which are newsletters, signatures and noise — files each one where it belongs, never files the same thing twice, replies to the sender when your rules say so, and asks a person whenever more than one answer looks plausible. Others are ready to go beside it: an invoice triage agent, a contract watcher, a data courier, a knowledge curator, and a company assistant that answers questions with citations and can put the rest of them to work.
Automation, or a colleague? A rule that moves every attachment into a folder is automation, and DocReveal has always been able to do that. What an agent adds is judgement: which attachment, which folder, is this the same document that arrived last week under a different name, is this sender behaving oddly, is this the moment to ask someone. Automation follows the path. A colleague decides which path it is.
Colleagues you can actually manage
Digital workers only scale if the organisation can hold them to account, and that is a governance question long before it is a technical one.
Every agent runs under your policies. Reading is usually free; anything that changes the outside world can be set to wait for a human, and the most sensitive categories always do. Approvals arrive in one inbox rather than in somebody’s chat window. Every step an agent takes — each decision, each capability it used, each answer it received — is written to a trail you can replay months later.
They also remember, which is what keeps them consistent and cheap to run. Each agent keeps a ledger, so the same message or document is never processed twice; a summary of what happened in earlier runs; and a set of notes and corrections that a person has confirmed. Correct an agent once and it stays corrected.
And when something is wrong, budgets and a kill switch stop it — per agent, immediately.
Bring the agents you already have
Any organisation of size already has something agentic running somewhere: in a copilot, on a cloud agent platform, inside one team’s own project. Agent Pool does not ask you to throw that away.
DocReveal speaks the open standards those platforms use, so the agents you already run can connect to your factory. They can call your DocReveal agents, use the same governed catalogue of capabilities and reach the same validated document truth — under the same policies, budgets and audit trail as everything else. What you have already invested in gains a memory and a governance layer instead of a rival.
What to take into your AI roadmap
Three questions worth asking before the next AI project starts:
- Who in your company is allowed to automate? If the honest answer is “engineering, when there is capacity”, your backlog is a strategy problem rather than a resourcing one.
- Could you prove what your AI did last quarter? Agents without an audit trail become a liability the moment somebody asks.
- Do your agents share anything? If each one keeps its own private understanding, you are running piecework — not a workforce.
The factory built the line, the warehouse and the control room. Agent Pool brings the people. Machines act, factories remember — and now the factory can hire.