An agent that finishes the jobAI agents for businesses and individuals
Custom AI agents that read, decide and act across your tools: they research a lead, update your CRM, draft the reply and ask you before anything important goes out.
WHAT IT DOESLOG
New lead from your site
→ Researched, scored, added to CRM
Maintenance request
→ Contractor booked within limits
Supplier invoice
→ Matched to the order, filed
Friday report
→ Pulled from 3 tools, sent to you
“Half your day is small jobs that need a bit of thought: look something up, check another system, update a record. None of it is hard. All of it pulls you away from the real work.”
SOUND FAMILIAR?
What is an AI agent for business?
An AI agent is software that works through a task the way a capable assistant would: it reads what came in, decides what needs doing, uses your tools to do it, and checks the result. Where an automation follows fixed steps, an agent can handle cases that are each a little different, such as a maintenance request, a supplier query or a new lead that needs researching. You set the goal, the rules and the limits. The agent does the legwork and asks you whenever a decision is yours to make.
Tamkoz designs and builds custom AI agents for businesses and individuals. We start with one job that eats your week, build the agent on tools you own, rehearse it on your real past cases, and switch it on with your approval on every action until you trust it. If a simpler automation or an off-the-shelf tool would do the job, we’ll say so.
THE JOBS IT TAKES ON
AI AGENTS
New lead from your site
→ Researched, scored, added to CRM
Maintenance request
→ Contractor booked within limits
Supplier invoice
→ Matched to the order, filed
Friday report
→ Pulled from 3 tools, sent to you
Anything unusual
→ Stopped and passed to you
Jobs an AI agent takes off your plate
If a job needs a few minutes of looking things up, a small decision and an update somewhere else, an agent can usually do it.
Requests that need checking in three systems before you can reply
→ The agent gathers the facts, drafts the reply and the next action, and waits for your OK.
New leads you never get time to research
→ Each lead is looked up, scored against your ideal customer and added to your CRM with notes.
A CRM that’s always out of date
→ Emails, calls and orders logged against the right customer automatically.
Invoices and orders that need matching
→ Documents compared line by line, with differences sent to you, not buried.
Chasing people for missing information
→ Polite reminders sent, replies read and the record updated when the answer arrives.
Weekly reports stitched together by hand
→ Figures pulled from each tool and a plain-English summary in your inbox on Friday.
One request, handled end to end
A tenant’s boiler breaks at breakfast time. The agent does the legwork; you make the one decision.
STEP 1
Request arrives
“The boiler at 14 Elm Court has stopped working and there’s no hot water.” Email from a tenant, 07:48.
STEP 2
Agent checks
Finds the tenancy, the landlord’s repair limit and the property’s usual heating engineer.
STEP 3
Agent arranges
Asks the engineer for slots, gets 13:00 today, drafts the work order within the landlord’s $320 limit.
STEP 4
You approve
One tap. Anything over the limit, or unusual, always waits for you.
STEP 5
Everyone told
Tenant and landlord updated, job logged in the property system, follow-up set for tomorrow.
SAMPLE WORKFLOW · A FICTIONAL LETTING AGENCY
Who uses AI agents?
Any business, or busy individual, with jobs that are too varied for simple automation and too routine to deserve your attention.
Letting and property managers
Maintenance requests triaged, contractors arranged within limits, tenants and landlords kept informed.
Agencies and consultancies
Leads researched, proposals started from a template, client updates drafted from project tools.
Online shops
Order problems investigated, supplier delays chased and customers updated before they ask.
Accountants and bookkeepers
Client documents requested, checked for what’s missing and filed in the right place.
Recruiters
Applications screened against your criteria, shortlists summarised, interviews scheduled.
Trades and field services
Job requests turned into booked visits with the right parts and notes ready.
Sole traders and individuals
Inbox, calendar, invoices and follow-ups kept in step without hiring an assistant.
A GOOD FIT IF…
- Your team jumps between the same apps all day
- Jobs need a little judgement, not just copying
- Simple automations have hit their limits
- You want a person to approve anything that matters
What we build
A job description for the agent
In writing: the goal, the steps a good human would take, what a good result looks like and what must never happen.
A working agent on your tools
Connected to your email, CRM, calendar, spreadsheets or other systems through their official connections.
Limits and checkpoints
Allowed actions, value limits and the moments it must ask you. Risky actions always wait for a person.
A log you can read
Every step the agent took and why, so you can check its work and spot problems early.
A rehearsal on your past cases
Before going live, the agent works through real examples from your history so we can measure how it does.
Monitoring and a plain-English guide
Alerts if it fails or gets stuck, and a guide to what it does and how to pause it. Everything in your name.
How we build an AI agent
One job, clear rules, a rehearsal on your history, then a careful go-live.
- ORDER 001START
1×Pick one job
We find a job that takes real time, needs a little judgement and happens often enough to matter, and write down exactly what a good result looks like.
- ORDER 002BUILD
1×Set the rules
Which tools the agent may use, what it may change, what it must never do, and the moments it has to stop and ask you.
- ORDER 003CHECK
1×Build and rehearse
We build the agent on your tools and run it against a few weeks of your real past cases before it touches anything live.
- ORDER 004GO LIVE
1×Go live, watched
It starts with your approval on every action. You see a log of what it did and why, and loosen the reins one step at a time.
What our agents connect to
An agent is only as useful as the tools it can reach. These are the ones we connect most often.
- AI MODELS
- OpenAI (ChatGPT models)Anthropic (Claude)Google (Gemini)
- AGENT AND WORKFLOW PLATFORMS
- n8nMakeZapierCustom services where needed
- EMAIL AND CALENDARS
- GmailMicrosoft OutlookGoogle Calendar
- CRMS AND HELP DESKS
- HubSpotPipedriveSalesforceZendeskFreshdesk
- FILES AND DATA
- Google DriveOneDriveSharePointGoogle SheetsAirtableNotion
- ACCOUNTS AND OPERATIONS
- XeroQuickBooksShopifyProperty and booking systems
If a system has no way for other software to connect, an agent can’t use it safely. We check this before you commit to anything.
How much does an AI agent cost?
Building the agent is a one-off job we quote in writing. Running it has two small monthly costs you pay directly: the AI model, charged by how much text it reads and writes, and any platform it runs on.
WHAT AFFECTS THE PRICE
How many steps and systems
An agent that reads email and updates one CRM is a small build. One that works across five tools is a bigger one.
How much judgement
Clear-cut decisions are quicker to build and test than ones that need careful rules and many examples.
Checkpoints and limits
Approval steps and value limits take a little extra work, and we always recommend them to start.
Volume
Hundreds of cases a day need a sturdier set-up, and more monitoring, than a few a week.
Model choice
Smaller, faster models cost far less to run. We use the smallest one that does the job well.
FROM THE PROVIDERS THEMSELVES
- AI models are priced per million tokens (a token is roughly three-quarters of a word). Anthropic’s smallest current Claude model costs $0.10 per million input tokens and $0.50 per million output tokens, for prompts up to 100,000 tokens. Source: Anthropic.
- OpenAI publishes per-token prices for each of its models, with smaller models costing a fraction of the largest ones. Source: OpenAI.
CHECKED 10 OCTOBER 2026. PROVIDERS CHANGE THEIR PRICES, SO CHECK BEFORE YOU BUY.
YOUR QUOTE · FIXED AND IN WRITING
We quote the build in writing, in your currency, and estimate the monthly running cost before you commit.
How long does it take?
A focused agent for one job can usually go live within a few weeks, and it earns its keep from the first week of supervised running.
The most common delay is access: logins and permissions for the tools the agent needs. We send a short list on day one.
TYPICAL TIMELINE
YOUR QUOTE HAS THE REAL DATES
01 DISCOVERY CALLONE CALL
We pick the job and walk through how you do it today.
02 QUOTE AND RULESA FEW DAYS
Scope, price, the agent’s rules and limits, in writing.
03 BUILD AND REHEARSETWO TO FOUR WEEKS
Built on your tools and tested against your real past cases.
04 SUPERVISED GO-LIVETWO WEEKS
Every action approved by you while we tune the instructions.
AI agent questions
Something else? Ask in the chat. A person answers.
What is an AI agent?
An AI agent is software that can work through a task in steps, deciding what to do next as it goes. Given a goal, such as dealing with a maintenance request, it reads the request, looks things up in your systems, takes actions like drafting an email or creating a job, and checks the result. A chatbot mostly answers; an agent gets things done.
How are AI agents used in business?
Common jobs are handling incoming requests (support, maintenance, orders), researching and qualifying leads, keeping a CRM up to date, reconciling documents against orders, preparing reports from several tools, and chasing missing information. The best first jobs are frequent, a bit fiddly, and easy to check.
What’s the difference between an AI agent and automation?
A normal automation follows fixed steps: when this happens, do that. An agent can choose its steps, which helps when every case is a little different. Agents cost more to build and run, so we use plain automation wherever it does the job, and an agent only where judgement is needed.
Is it safe to let an AI agent act on its own?
Only within limits you set. Every agent we build has a list of allowed actions, spending or value limits, and checkpoints where it must ask you. It starts with your approval on every action, keeps a log of what it did and why, and can be paused at any time.
Do I need a custom AI agent, or will an off-the-shelf one do?
If a ready-made agent in a tool you already use (your help desk, CRM or email) covers the job, use it: it’s cheaper. Custom agents make sense when the job crosses several tools, follows your own rules, or needs to work with a system nothing else connects to. We’ll tell you which applies.
Which tools and models do you build agents with?
We pick for the job: models from OpenAI, Anthropic or Google, and platforms such as n8n, Make or a small custom service when a job needs more control. Your agent connects to your apps through their official connections, and the accounts are set up in your name.
What happens when the agent gets something wrong?
It will, sometimes, which is why the rules matter. Anything it isn’t confident about goes to you, risky actions always need approval, and every action is logged so a mistake can be found and undone. We review the log with you in the first weeks and tighten the instructions.
Can individuals and sole traders use AI agents?
Yes. A one-person business often gains the most, because there’s nobody else to hand the fiddly jobs to. An agent that keeps your inbox, calendar and invoices in step can give a sole trader hours back each week.
How long does it take to build an AI agent?
A focused agent for one job usually takes a few weeks from first call to going live, including the rehearsal on your past cases. Agents that work across many systems take longer and go live in stages.
Further reading
- AI agent examples: 12 real jobs, and 5 to keep for yourselfTwelve jobs agents do well today, the checkpoint each one needs, and the jobs to keep away from them.
- Which business tasks should you automate first?The same scoring method helps decide whether a job needs an agent, a simple automation or neither.
GOES WELL WITH
- AI automationWe connect the tools you already use so the repetitive admin (quotes, follow-ups, invoices, data entry) happens on its own.
- RAG & AI knowledge basesA private AI assistant that answers from your own documents, policies and data, shows where each answer came from, and keeps your information yours.
- AI integrationWe connect models like ChatGPT, Claude and Gemini to your website, CRM, app or internal systems, so AI works on your data, inside your workflow, under your control.
SOURCES · CHECKED 10 OCTOBER 2026
Give the fiddly jobs to an agent.
Tell us the job that eats your week. We’ll tell you honestly whether an agent, a simple automation or neither is the answer, and quote in writing.