Skip to main content

AI sales agents

Find customers without spending your partners’ time on it.

Finding customers breaks down into four channels, and each into steps that do not swap round. Enriching a list before the target has been framed means paying for data for nothing.

Families of AI agents, not an off-the-shelf catalogue

What follows describes the families of AI agents we build and the way we build them. It is not a shelf of ready-made products: every agent is tuned to your data, your rules and your exceptions. We publish no agent as deployed until a client has allowed us to tell the story.

01

Outbound

You want to create meetings without waiting for inbound.

The whole chain, from framing the target to handling the replies. The order is not negotiable: frame, source, enrich, then write.

  1. 1.1

    Sourcing and enrichment

    • ICP setupQuestions the client to define their target, then turns what they say into precise criteria the following AI agents use.
    • Company sourcingFinds the companies that match the client's criteria, removes duplicates and writes them into the CRM once approved.
    • Signal detectionWatches the events that create a need (funding round, hiring, new technology) and pulls out the companies concerned.
    • Sourcing from signalsStarts from companies that have just triggered a signal, completes their record and writes them into the CRM with the context.
    • Decision-maker sourcingIdentifies, in each target company, the people to contact, attaches them to their company and writes them into the CRM.
    • Buying committee mappingOn a large account, spots everyone who carries weight in the decision, gives each a role and flags the roles not covered.
    • EnrichmentRetrieves verified contact details for decision-makers, going from the cheapest provider to the most expensive, and stops as soon as it finds one.
    • Volume prioritizationScores each lead on how well it fits the target, ranks the list from best to worst and cuts it into waves.
  2. 1.2

    Campaign

    • Campaign buildWrites the message sequence (first contact and follow-ups) for each decision-maker, with three levels of personalization.
    • Multi-contact orchestrationCoordinates the messages sent to the different decision-makers in one account: who gets what, when, and in what order.
    • Warm paths inLooks for a shared connection who can introduce the target decision-maker, then drafts the introduction request.
    • Contact loadingLoads the ready contacts into the sending tool, links them to the right sequence and leaves them paused until the green light.
    • Deliverability and ramp-upSpreads sending across several mailboxes, raises the volume gradually, holds the daily caps and watches the reputation.
  3. 1.3

    Replies and steering

    • SynchronizationPulls back what happened in the sending tool (sent, opened, replied, bounced) and updates each lead's status in the CRM.
    • Reply handlingSorts the replies received, spots the warm ones and prepares a draft answer to be approved before sending.
    • ReportingGathers the campaign figures, checks that the emails arrived and sends a written summary to the client.

What this channel does not doThis channel touches people and their professional data. The notices, the right to object and the retention periods are framed before the first send, not after.

02

Partnerships

Your growth runs through referrers, resellers or integrations.

The most profitable channel and the least equipped. The same engine as outbound, tuned to partners instead of end customers.

  1. 2.1

    Partner sourcing

    • Partner mappingMaps the possible partners around the client (agencies, tools, integrators) and ranks them by how well they complement you.
    • Partner decision-makersFinds, at each target partner, the person who runs partnerships and attaches them to their company.
    • Synergy caseSpells out, for each partner, what each side gains from working together, a possible joint offer and a simple first step.
  2. 2.2

    Shared engine

    • Shared engine, partner modeReuses the prospecting chain tuned to partners: same steps, same guardrails, different target.
  3. 2.3

    Activation

    • IntroductionsCross-references the partner's network with the target accounts, proposes the useful introductions and drafts the request.
    • Referral programmeSets up the referral programme: terms, commission scale, tracking link and welcome pack.
    • Referral and commission trackingLinks each signed deal to the referrer who recommended it, calculates the commission due and prepares the statement.
    • Reseller enablementPrepares the reseller's sales kit (pitch, FAQ, pricing, objections) and their onboarding path.
    • Deal registrationRegisters the deals declared by resellers, spots duplicates on the same prospect and tracks deals in progress.
    • Integration framingFrames the technical integration before any development: use cases, feasibility, scope of the first version.
    • Co-building and listingFollows the build of the integration, writes the documentation, prepares the listing page and the joint announcement.
    • Co-marketing planBuilds the joint action plan with a partner: formats, calendar, roles and target audiences on both sides.
    • Co-production and distributionProduces the joint content with the partner, distributes it to both audiences and measures the reach.

What this channel does not doA partnership is signed between two leadership teams. The AI agents prepare, qualify and track. They do not negotiate and they do not replace the relationship.

03

SEO

You want to be found on the queries that bring customers.

From keyword strategy to production, with a quality gate that blocks thin pages. Producing a thousand pages with no value degrades a site instead of serving it.

  1. 3.1

    Strategy

    • SEO strategyFrames the target and the goals, splits the subjects into pillars and satellites, and sets the publishing pace.
    • KeywordsSpots the keywords already close to page one, finds new ones, then ranks them by priority.
  2. 3.2

    Content

    • DifferentiationAnalyses the competitors, compares the offers point by point and gathers the sourced figures that give you an angle.
    • WritingWrites a full article from a keyword, links it to the neighbouring pages and leaves it as a draft.
    • Programmatic page generationGenerates a batch of pages at once from a dataset and a template, each with its own content.
    • Quality gate on generated pagesReviews the generated pages, spots the ones that are too thin or too alike, and decides what happens to them.
    • Local pagesWrites one page per town or area, with content specific to the place and reviews, then links it to the rest of the site.
  3. 3.3

    Authority and steering

    • Inbound linksSpots the sites, podcasts and directories that could cite the brand, prepares the requests and tracks the links obtained.
    • Business listing and citationsCompletes the business listing and aligns the name, address and phone number across every directory.
    • SEO reportsGathers the week's figures, works out the visitor journey, spots the pages that are falling and the easy wins.

What this channel does not doSEO is measured in months, not weeks. We do not sell a guaranteed position, and nobody can.

04

Answers from AI

Your prospects ask an AI assistant before they open a search engine.

Being quotable by the assistants, and knowing whether you are. A young channel, where the ground is taken now.

  1. 4.1

    Quotability

    • Quotable factsTurns the company's figures into short, sourced facts that AI assistants can quote as they are.
    • Formatting for AIAdds a short summary at the top of the page, extracts that stand on their own, an FAQ and the markup assistants can read.
  2. 4.2

    Watch

    • Citation trackingAsks the AI assistants your customers' questions, notes whether your brand is cited and tracks the gap with your competitors.

What this channel does not doThe assistants change their answers without notice and without a log. We measure a trend on a sample of queries, never a position.

An AI agent, concretely

What an AI agent does, step by step.

An example on a bid. Four steps, and a human who decides at the last one. No client runs this exact setup: it is an example, to show what we are talking about.

An example, not a setup installed at a client
  1. Automatic

    It reads

    It opens the tender documents and pulls out the list of required paperwork, the scoring criteria and the dates.

    • Shared mailbox
    • SharePoint
    • The tender PDF
  2. Automatic

    It gathers

    It finds, in your past bids, the paragraphs, the project references and the CVs that match the criteria.

    • Your past bids
    • Your references
    • Your CVs
  3. Automatic

    It drafts

    It assembles the draft response in your template, and flags the documents that are missing or out of date.

    • Word
    • Your response template
  4. Bid owner

    You approve

    The bid owner reviews, corrects whatever they want, then sends. Nothing goes out without that step.

    • An approval screen
    • Outlook

The time is saved on the first three. The fourth never disappears: it is your signature.

End to end

Five links. Most deliver only one.

An AI agent is not a project. It is the middle link. Before it comes the data, after it come a screen and monitoring, otherwise nothing runs beyond the demo.

  1. Collection

    Go and get the data where it lives.

    Your tools, your documents, your shared mailboxes, and public sources where they exist. Data that has to be retyped by hand to feed an AI agent does not last three weeks.

  2. Structuring

    Clean it and reconcile it before using it.

    Deduplication, one common format, matching by identifier. The rule we never work around: we match on an identifier or a domain, never on a name, because namesakes manufacture errors that make no noise.

  3. The AI agent

    The work itself, with a closed scope.

    Every AI agent says what it does and what it does not do, and points to the next one for the rest. An agent that does everything is an agent you can neither test nor correct.

  4. The application

    The screen your teams use it from.

    An AI agent with no interface stays a script only its author knows how to launch. Your teams have to be able to trigger it, see what it proposes, correct it and approve it, from a tool they already open.

  5. Monitoring

    Know what is running, what it costs, and how to stop it.

    A log per run, the cost per AI agent, and a documented rollback before go-live. That is the difference between a service and a demo.

Our build rules

What an AI agent of ours always does.

Five rules we do not negotiate. They can be checked in the code and the documentation we hand over, not taken on trust.

  • The AI agent stops and shows

    Before writing into your tools or sending anything outside, the AI agent presents what it has prepared and waits for your approval. The stopping point is written into the specification, not left to a setting.

  • Nothing hard-coded

    No identifier, no column, no criterion frozen into the code. Everything goes through a readable configuration your teams can change without us.

  • A closed scope

    Every AI agent carries the list of what it does not do, and the name of the one that does. That is what makes the system testable, and its cost predictable.

  • Matching by identifier

    Never by name. Two companies with the same name, or two spellings of one client, produce errors that stay invisible for months.

  • A documented rollback

    Written and tested before opening, not improvised on the day it jams. You can go back to the previous state at every step.

Contact

Which channel deserves your next euro?

That is the question the diagnostic answers. Three weeks, and you know which one to open first, with its costing.

Tell us about your project