Your own CoDNA
A private knowledge base for your company, hosted by us or on your own servers, with what you already keep in Confluence, Notion or Google Drive brought in.
Services
We bring AI into your company, and it keeps working after we leave.
We come in, set up CoDNA with your documents and meetings, connect Claude and ChatGPT to it, automate two or three of the tasks your team repeats every week, and teach people to use it on their own work. It takes eight weeks, and we measure the result against where you started.
Book a 30-minute call How a project runs
Also in Georgian: ქართულად
What you get
AI that knows your company
Claude, ChatGPT, Claude Code and Codex connect to it. Each person signs in as themselves, and their AI sees only what they can see.
Work that runs itself
Two or three recurring tasks automated, such as meeting notes turned into decisions, weekly reports, or answers to the questions people keep asking. A person approves each result.
A team that carries on
Short sessions on your own work, and two to four champions from your team who can build the next automation without us.
What changes for your business
Six examples of what a project changes day to day. In your project we pick the two or three that matter most to you, and measure them.
Meetings that leave a record
Before: decisions from the weekly meeting live in a few people's notes, and a month later nobody is sure who agreed to what.
After: the meeting is transcribed in Georgian, an agent drafts the decisions and action items, and the meeting's owner approves them. Anyone can ask "what did we decide about this?" and get the answer with a link to the meeting.
New people up to speed sooner
Before: a new colleague spends the first weeks asking around where things are and how they're done.
After: an onboarding assistant answers from your own documents, with a link to the source, so experienced colleagues are interrupted less.
Ready answers to repeat questions
Before: the same questions about procedures, prices or policies take up someone's time every day.
After: AI drafts the answer from your approved pages, and a person checks it before it goes out.
Weekly reports, drafted for you
Before: someone spends Friday afternoon collecting updates for the weekly report.
After: a draft is ready on Friday morning, built from the week's tasks, meetings and changes, and that person only has to check it.
Code that follows your rules
Before: reviewers repeat the same comments, and the conventions live in senior engineers' heads.
After: AI checks every pull request against your written conventions before a person looks, and coding agents read your architecture notes.
Knowledge that stays when people leave
Before: when someone leaves, what they knew leaves with them.
After: decisions, procedures and runbooks get written down as part of the daily work, where people and their AI tools can find them.
What you gain: less time on repeat work, faster answers you can trust, and a company memory your AI tools can use. We agree the measures in the first week, so you see the difference in numbers.
Why it works here
- Meetings in Georgian. Teams and Zoom don't caption Georgian. A meeting held in CoDNA records each person on their own track, so the transcript says who said what, and its decisions end up in your wiki.
- Your data in Tbilisi. Hosted on our own servers in Cloud9's data centre in Tbilisi, or on yours. Meeting audio is transcribed by Google's Gemini.
- Results you can check. In the first week we agree three goals and measure where you start. At the end we report against those numbers.
- Nothing locked in. Every page is plain markdown, each person can download what they can read at any time, and what we build stays with you.
How a project runs
Five steps over eight weeks. You sign off each one as it ends, and from week three they overlap, so people learn on the automations they will use.
- Discover, week 1. We talk to five to eight of your people, map the work that repeats, and agree three goals with their starting numbers.
- Set up, weeks 2 and 3. Your CoDNA, sign-in with your company accounts, your existing documents brought in, and your AI tools connected.
- Train, weeks 3 to 6. Short sessions by role, on your own work, and two to four champions picked from your team.
- Automate, weeks 4 to 7. We build the workflows you chose together with your champions, and measure each one.
- Hand over, week 8. A results report against the starting numbers, a runbook for each automation, and your AI use policy.
Three ways to start
Each has a fixed price, agreed before we start. If you go on within two months, what you paid counts toward the next step: the workshop toward the pilot or the program, and the pilot toward the program. We take on one new company at a time.
Workshop
One day on your team's real work, with your own code and documents. We start from what matters most to your business, build one real change with your team, and end the day by deciding together whether a pilot is worth running.
- Up to 12 people
- The first three training sessions, condensed
- A CoDNA to try for 30 days
Pilot
Three weeks. One team and one workflow, judged on one measure we agree at the start.
- Discovery, and one measure agreed with you
- Your CoDNA set up and connected
- One workflow automated, with a go/no-go report at the end
Program
Eight weeks. Everything above, across your teams, with the results measured.
- Two or three workflows automated
- Training by role, and champions
- A results report and runbooks
Who you would work with
Tornike Razmadze, who builds CoDNA, leads each project. We work this way ourselves: CoDNA is developed with Claude Code reading a CoDNA of the project's own decisions, and Medspace keeps its company knowledge in CoDNA too. Contracts are with Radium 2023 LLC in Tbilisi.
Start with a 30-minute call
Tell us what your team does and where the time goes. We'll tell you honestly whether AI will help, and where to start. The call is free.