Skip to content

AI agents that do the work, with you in control

An agent reads, decides and acts across your tools: it sorts email, updates records and drafts replies, then waits for approval where it matters.

Custom AI agents

Starting price
$5,000
Typical timeline
6
weeks to launch

Works with

Op

Work that rules alone cannot handle

Every message is different

Simple automations need fixed rules. Real emails, requests and documents rarely follow them.

Judgment calls pile up

Someone has to read each request, decide what it is and route it, dozens of times a day.

Busy people do low-value reading

Senior staff spend hours reading and summarising before they can make a decision.

How an agent handles a request

The agent prepares everything. A person approves before anything leaves the building.

  1. Request arrives

    Email or form

  2. Agent reads

    Understands intent

  3. Gathers data

    CRM, Sheets, Drive

  4. Needs approval

    Person reviews

  5. Action done

    Logged and sent

What we build for you

Email and inbox triage

The agent reads incoming email, labels it, drafts a reply and routes anything urgent to a person.

CRM and record updates

It pulls details from emails, calls and documents and updates the right records.

Document processing

Invoices, contracts and forms are read, checked and turned into structured data.

Research and preparation

Before a sales call, it gathers the company background and history and writes a short brief.

Multi-step tasks

It works through a task across several tools, such as checking stock, drafting a quote and logging it.

Approval checkpoints

Any action that sends, pays or deletes waits for a person to approve it.

Example: quote requests for a logistics firm

Before

A coordinator read each quote request, checked rates in a spreadsheet and typed a reply. Each one took about 20 minutes.

After

An agent reads the request, checks the rate sheet and drafts the quote. The coordinator reviews and sends it in under 2 minutes.

An illustrative example of a typical project.

Safety and accuracy, built in

AI is only useful if you can trust it. These rules are part of every build, not optional extras.

Least access

The agent only gets the permissions its job needs. It cannot reach anything else.

Human approval

Sending, paying, deleting and other sensitive actions always wait for a person.

Full activity log

Every decision is recorded with its reasoning, so you can check exactly what happened.

Shadow mode first

The agent proves itself on suggestions before it is allowed to act.

Tools we connect

All integrations
  • OpOpenAI
  • Anthropic
  • Gmail
  • Google Sheets
  • HubSpot
  • n8n
  • Make
  • Supabase

What you receive

  • A written scope of what the agent may and may not do
  • The working agent connected to your tools
  • Approval steps for every sensitive action
  • A log of every decision and action the agent takes
  • Handover guide and 30 days of monitoring

Timeline

  1. ScopeWeek 1

    We pick one job, define the agent's permissions and agree where approvals are required.

  2. BuildWeeks 2 to 4

    We build the agent, connect its tools and test it on your past examples.

  3. ShadowWeek 5

    The agent runs alongside your team, suggesting actions without taking them.

  4. LaunchWeek 6

    We turn on approved actions step by step and review its work weekly.

Custom AI agents

From $5,000

for one agent with approval steps and logging. AI usage is billed at cost.

Questions we hear often

How is an agent different from an automation?

An automation follows fixed rules. An agent can read unstructured information, decide what it means and choose the next step, within limits you set.

Can the agent make mistakes?

Yes, which is why we use approval steps, shadow mode and logs. Sensitive actions never happen without a person.

Where does the agent run?

In your own accounts or a private server, using tools like n8n or a small custom app. You own it.

What does it cost to run?

AI usage for most agents costs between $30 and $300 a month, depending on volume. We estimate it during scoping.

Find the first workflow worth automating

A 20-minute call is enough. We tell you plainly what we would automate first, and what it would cost.

What happens on the call

  1. 1We look at how the work happens today
  2. 2We name the first workflow worth automating
  3. 3You get a fixed written quote. No obligation

Free, 20 minutes, on Google Meet.