AI agents
Using ChatGPT doesn’t make a company AI-powered. Permanent AI agents do: they hold the front line, work inside your CRM and hand over to people with the full context.
We build that set of agents for you and keep it running — solutions that don’t need supervision. Today a lead comes in on Instagram at 11 pm, a phone call lands while the seller is in a meeting, a web form sits in an inbox until Monday. Each of those is a client who may already be talking to a competitor. Agents close that gap. They answer first, they never go on holiday, and they write everything down where your team can see it.
For whom
Companies of 30–300 people with a steady flow of inbound requests across phone, messengers, social media, email and web forms. Owners who still get pinged about every lead. Sales teams that lose deals on evenings, weekends and holidays.
The front line
Supporting agents
Joins the call under its own name, with consent, transcribes it, and turns it into a brief and proposed tasks where every line points at the transcript.
Our founders built it for: Consulting firms on Teams, Meet and Zoom.
See the case: Meeting AgentWrites the weekly digest, computes every deadline, posts colour-coded alerts, and flags anomalies for a person to decide on.
Our founders built it for: An advisory firm, a telecom operator and a payments platform.
Answers staff and client questions from your documents only, with citations, by chat or by voice; questions it cannot answer queue for the owner.
Our founders built it for: An energy company, an industrial training provider and a software product’s help centre.
See the case: Knowledge AgentTurns a voice note or a recording into a structured draft record that a person keeps, edits or discards.
Our founders built it for: Two healthcare platforms and an advisory firm’s partners.
Watches a shared drive or inbox, classifies each file, extracts the fields with a confidence score, and lets staff confirm or override.
Our founders built it for: A law firm, an advisory firm and a payments platform.
See the case: Document Intake AgentTurns a user’s bug report or complaint into a structured task for the right team.
Our founders built it for: Three software products.
Built by our founders in previous roles; each runs inside the client’s own systems. Client names are under NDA.
Start here
We never automate chaos. The sprint starts by describing and costing the process; then we build it into your systems, secure it and measure the result in hours and money.
We describe the process — trigger, inputs, steps, output, errors — calculate what it costs you today, and agree the one number that decides go or no-go.
One AI agent or automation, wired to your CRM and your rules. It runs in shadow mode first.
Security review before launch. Then it runs in production, with a person able to step in at any point.
A before-and-after in hours and money, documentation, and the code, prompts and data — all yours.
How it works
Start small with the pilot sprint above. The full front line follows this plan.
We list every touchpoint, every lead category, who should own which deals, and what “typical request” means in your business. Deliverable: the routing and pricing rules the agents will follow, signed off by you.
Front Desk Agent live on all channels, wired to your CRM. Deliverable: each inbound lead answered, enriched, categorized and assigned, with first-response time measured before and after.
Quote & Invoice Agent and Owner’s Agent go live. Deliverable: typical deals closed without a human; the owner steering by voice.
We watch what the agents handled and what they escalated, and tune the rules. Deliverable: a monthly report in numbers.
Proof · founders’ previous roles
Agents our founders built in previous roles. Client names are under NDA; the numbers are not.
What we ship by default
The agent steps back, the person steps in, the client notices nothing. Inside the rules you signed off, the agent completes the request; outside them, or below its confidence threshold, it drafts and a person sends. Every correction a person makes is logged next to the suggestion.
Every message, quote and invoice the agent produces is logged where your team already works.
Routing, pricing, who may talk to whom. If a rule does not exist yet, we write it with you in the first two weeks. The model never prices, scores or sets a rate: calculations are deterministic and cited, the model writes the words. An agent closes only what your rules allow, such as a typical request at a listed price; anything else goes to a person with a one-paragraph brief.
Private deployment, model-agnostic (OpenAI, Anthropic, Google, open-weight models), on API terms that exclude training on your data, with zero data retention where the provider offers it. No consent, no model call; a model receives only the data the task needs, and the rest is masked.
Handled, escalated, missed. Every model call is accounted for by cost and latency, with a daily spend cap as a hard stop. A new agent runs in shadow mode before it goes live. If an agent is not paying back, we rework it or switch it off.
Email, WhatsApp (with voice notes), Telegram, website chat and voice, phone callbacks, LinkedIn, Slack, and meetings on Teams, Meet and Zoom.
Your CRM, accounting, e-signature, bookkeeping, shared-drive document stores and automation platforms. Our founders have connected all of these before.
FAQ
We quote after the intro call, once we know the process. A pilot sprint covers one process, with the scope agreed before we start. Running costs are reported per model call, with a daily spend cap as a hard stop.
We agree one success metric before the build. A new agent runs in shadow mode on your real requests before any client sees it, and we measure it every month after launch. If it isn’t paying back, we rework it or switch it off.
Yes. Each agent tells clients it is an AI assistant at the start of the conversation, and any client can ask for a person at any time.
The Front Desk Agent answers the phone, takes the request, and either resolves it or books a call with the right person. Voice is a first-class channel, not an add-on.
Yours. We integrate with what you already use; if you have none, we set one up as part of Launch.
Yes. Agents our founders built in previous roles run today inside clients’ own systems in tax and legal advisory, insurance, payments, consulting, healthcare, energy and telecom. Five are described on the Case Studies page.
Agents show what clients ask and where deals stall; LOGITRON Analytics shows how your messages are understood and where decisions are distorted. Together they show both sides of the front line.