
Agency OS: why I built my own Grok Bot alternative
An agency OS puts your AI agents, client data and way of working in one system you own. Why Grok Bot wasn't enough, and how I built my own.
I help web agencies run on agentic AI.
I’ve run an agency for 20 years and spent the last three changing how agencies work with AI, starting with four of my own. Now I work inside other agencies as their AI lead, helping owners and their teams build differently, safely.
WHERE MOST AGENCIES ARE
Everyone has a chat window open, each on a different tool, and the work still gets done the old way. Agentic AI is different. You give an agent a job, not a question, and it does the work while your people direct it and sign it off. I made that change in my own agencies first.
AGENCIES I WORK WITH
WHY WORK WITH ME
I founded Code23 in 2005 and I’ve run it through the boom years and the hard ones. I know how an agency’s margins, deadlines and teams feel from the inside, because I still live with them.
Four of them, rebuilt around AI before I asked anyone else to change.
The agents, build loops and QA set-up I bring are the ones I use every day.
No account managers, no junior consultants, no hand-off after the sale.
Your team makes the calls that matter and owns the result. We agree which tools touch client data, and nothing goes live without your sign-off.
WHAT I DO INSIDE YOUR AGENCY
Decide where AI earns its place in your agency: in delivery, in the back office and in what you sell to clients. Tool choices, standards, and how you’ll know it’s working.
PRIORITIES / TOOL DECISIONS / WORKING STANDARDS
Your agency’s own standards written into its agents. Build loops that check each piece as it’s made. QA agents that test like a real user before anything goes live.
AGENT STANDARDS / BUILD LOOPS / QA AGENTS
Find the champion, start them on the job they’re sick of doing, and let everyone else see it working. It sticks when the team owns it.
AI CHAMPIONS / HANDS-ON SESSIONS / ONGOING SUPPORT
HOW THE WORK CHANGES
Take a familiar piece of agency work. Agents take on the volume in between. Your team makes the calls that matter.
Client context, goals and the work to be done.
Research, drafts and a structured start, built by agents to your standards.
Your team shapes it, challenges it and makes the calls.
Automated checks and a QA agent go over it before your team signs it off.
HOW IT WORKS
Month one is the strategy on every plan, at the plan’s price. Advisor starts with a kick-off and a one-page AI plan. Partner and Lead start with the full map of your agency and a 90-day plan. You keep the plan whatever you decide next.
£750 to £3,000 a month (+VAT).
Advisor £750, Partner £1,500 or Lead £3,000 a month. Month to month, with no long-term commitment. Need me one day a week or more? Let’s talk.
Each month we add the next agent and sharpen the last. You own everything we build.
ALL PRICES (+VAT).
A FEW QUESTIONS

An agency OS puts your AI agents, client data and way of working in one system you own. Why Grok Bot wasn't enough, and how I built my own.

AI agent memory, explained: one shared record of what your agency knows that every agent reads first. What goes in it, what stays out, and how to start one.

A fractional chief AI officer leads AI in your business for a few days a month. What the role covers, how it differs from a hire, and whether you need one.