The TrueValue Agent

AI Agent for M&A Teams

Not a chat box waiting to be asked. The TrueValue Agent reads what lands on your deals overnight, measures it against your mandate, prices what it can from the accounts, chases what is outstanding, and brings you decisions with the evidence attached. You make the decision. TrueValue does the groundwork.

Start my 14-day free trial Watch the Agent analyse a deal
  • £399/month
  • Unlimited users and deals
  • TrueValue Agent included
  • No credit card required

What an AI agent for M&A is

An AI M&A agent is software that works a deal on its own initiative, within limits a team sets: it reads new documents and messages as they arrive, runs the analyses the team would run, watches for the signals that a deal is moving or drifting, and proposes — or, where permitted, takes — the next action. The distinction from an AI assistant is initiative. An assistant answers when asked. An agent has a job.

The TrueValue Agent lives inside the TrueValue workspace, which is what makes it useful rather than merely clever: it reads the same deal record, the same extracted accounts and the same diligence list the team does, and it acts through the same tools. A finding it files is on the deal, with the records it was read from named, and a proposal it makes can be approved with one click.

What the Agent does across the workflow

  • Screens what arrives against your mandate

    New listings from the marketplaces, inbound CIMs and companies from the register are measured against the acquisition mandate you set as numbers. The ones that clear the bar are the ones you read.

    How to build an acquisition mandate

  • Analyses CIMs and extracts financials

    A memorandum dropped on a deal is read, spread and scored, with the red flags and the missing information listed and every figure tied to its source.

    AI CIM analysis

  • Prices a target before you ask it to

    When accounts land, the Agent runs the five-method valuation and the returns model — IRR, cash cover and the maximum price at your hurdle rate — and says when an assumption has drifted from the latest filed figures.

    M&A valuation software

  • Identifies risks

    Customer concentration, owner dependence, adjusted earnings that outrun reported earnings, a covenant that a growing working-capital need would breach. Each is a finding with the evidence and a severity.

  • Runs the diligence list and reads the replies

    Outstanding requests are chased on the cadence you set, responses are filed against the item they answer, and red flags found in what comes back are surfaced.

    M&A due diligence software

  • Surfaces changes and monitors the deal

    Reply times, data-room attention, silence after a request and which buyers are real are read as one picture per deal, so a drifting one reaches you while there is still something to do about it.

  • Recommends next actions

    Each finding arrives with a proposed next step — a chase email, a stage move, a request list, a memo — that you approve, edit or dismiss in one click.

  • Drafts the documents

    IC memos headed by a deterministic fact pack, teasers and information memoranda from the deal record, outreach in your firm's voice. A draft, never a send.

    M&A document management

How a run works

The Agent runs on a schedule, on events, and on demand. Each run is the same loop.

  1. Step 1

    Read

    It reads the deal record through the same tools a person uses: the pipeline, the accounts, the diligence list, the inbox, the data room, the register. Counterparty text is treated as data, never as instructions.

  2. Step 2

    Judge

    Momentum, mandate fit, valuation, returns, covenant tests and readiness are computed by deterministic engines. The model reads figures; it does not produce them.

  3. Step 3

    Propose

    A finding names its evidence and the action it proposes. A figure with no source in the record is refused, not repaired.

  4. Step 4

    Act, within limits

    You set an autonomy level per kind of action: ask first, tell you after, or get on with it. Anything that leaves the workspace always asks. Every action leaves a receipt with an undo.

The guardrails, as enforced

These are rules the runtime holds, not reassurance written for this page.

  • It shows its work

    Every finding names the records it was read from, and every run leaves a step-by-step trace you can open.

  • It cannot invent a figure

    Every number the Agent writes is checked against the figures in its own tool results. A figure with no source is refused, and the refusal is recorded.

  • You set what it may do alone

    Twelve classes of action, each with an autonomy level of ask, notify or auto, and a ceiling in code that a setting cannot raise. Outward actions — email, signature requests, diligence requests — are always ask.

  • Everything is reversible

    Each action writes a receipt carrying its inverse. Undo is one click, and it works even if the Agent is later switched off.

  • It is measured

    Outcomes are recorded on every action — replied, stage moved, document arrived, no change — and a per-class scorecard shows precision and effectiveness. Autonomy is only ever offered upward on measured results.

  • It asks

    Where the record cannot settle a question, the Agent asks it, with options, rather than guessing. Nothing executes on an answer.

Humans remain responsible for the decision

An investment decision is a judgement about the future made with incomplete information, and it belongs to the people who carry the consequences. The Agent is built to make that judgement better informed and to cost less time, not to make it. It does not submit an offer, sign anything, contact a seller or move a deal to closed without a person choosing to.

That is also why its output is designed to be checked. A committee reading an IC memo the Agent drafted sees every figure cited to its source, sees what the record does not state, and sees a section on what would make the recommendation wrong. The point is not to trust the Agent; it is to be able to verify it in the time it saves.

Three ways to work with it

  • The morning review

    Overnight, the Agent reviews every live deal and files what it found: N findings, M actions taken, K needing you. The dashboard opens on that line.

  • Missions and jobs

    Standing briefs on a cadence or on an event — a CIM arriving, an offer received, a counterparty going quiet — and longer jobs that own an outcome, such as running a sale process or working an acquisition through to a decision.

  • The conversation

    Ask it anything about a deal in plain English from any page. It reads the record, runs the engines and can approve, dismiss or undo its own proposals when you tell it to. The same conversation continues on the dashboard and in the side panel.

Frequently asked questions

Is the TrueValue Agent available now?

Yes. It is live and included in the TrueValue plan. Agent access is switched on for a workspace by the TrueValue team as part of onboarding, and every agent surface appears once it is.

What can it do without asking me?

Only what you have permitted, per kind of action, and only internal, reversible things: logging an activity, creating a task, filing a red flag, recording an economics scenario. Sending anything to anyone outside the workspace always waits for a click.

Can it make an investment decision?

No. It screens, analyses, prices, monitors and recommends. Investment and transaction decisions are made by people, and the runtime does not give the Agent a tool that could make one.

How does it avoid hallucinated numbers?

Two ways. Every figure in the product is produced by a deterministic engine or read from a document, and the AI writes narrative around them without being able to change one. And every number the Agent writes into a finding, a proposal or a memo is checked against the figures in the tool results it actually read; an unsourced figure is refused.

What does it cost?

Nothing beyond the plan. The Agent is included at £399 a month with unlimited users and deals. A daily safety ceiling on AI spend exists per workspace, set well above what real use produces.

Can I use my own AI assistant with TrueValue?

Yes. TrueValue exposes its tools over the Model Context Protocol, so Claude and other MCP clients can read and act on your workspace with the same permissions and audit trail.

Go deeper

Watch the Agent analyse a deal

The demo workspace is the real product on sample data. Open it, drop a CIM on a deal, and watch the first pass come back with the evidence attached.

Watch the Agent analyse a deal Start my 14-day free trial