4 hours → 20 minutes
A brief goes in. A branded, on-message first draft comes out, with your pricing and your terms already in it. Your best person only has to review it.
Increase Productivity Secure your Data Own your AI Stack Reduce Token Costs
Odokai puts a workforce of AI agents behind each of your people, working on your own documents, in your cloud or ours. From £500 a month (ex VAT), a fraction of the cost of a junior hire.
- Increase ProductivityConnect your data and create unlimited agents to do the work of a larger team
- Secure your DataKeep all your valuable data in your domain. No leaking out to training models
- Own your AI StackBrand it as yours and make it your own bespoke software suite, doing what hundreds of SaaS subscriptions do today
- Reduce Token CostsLet Odokai choose the best LLM for each task and cut your model spend drastically
The product
Agents that work on your own documents.
Ask in plain English. Agents read your documents, run the numbers and draft the reply. Anything that goes out waits for approval, and every step is logged.
Answers you can put in front of the board.
Ask a question of your pipeline, your ledger or last quarter's pack. The answer comes back as figures, charts and tables.
- Reads your CRM export, spreadsheets and documents
- Charts and tables drawn in the reply
- One click to the next step
Two hundred contracts, checked before lunch.
Agents read every document against your policy, flag what needs a person and propose the wording. Your experts review the flags.
- Every finding links to its page and clause
- Proposed changes shown as a redline
- Recorded against the policy version
Workflows that run on a schedule.
Workflows branch, loop over every item and pause for approval. Watch each step live, or come back to the result.
- Branches, loops and approvals on one canvas
- Retries and error routes built in
- Runs on a schedule or when something happens
Decisions wait on Home.
Home shows what needs you, what ran while you were away and what is coming next. Anything that goes out waits for approval.
- Approvals with the context to decide
- Your apps and boards one click away
- Everything that ran, on the record
Increase productivity
What agents take on.
Drafting, first-pass checks, reports and creative work, done on your own data and reviewed by your team. See ten jobs in detail.
Half a day → 15 minutes
Numbers pulled from every system, formatted, and checked against last month before anyone opens it. Someone reads it, signs it, sends it.
All morning → Minutes
First-pass review flags the risk clauses, pulls the key dates and figures, and lists what is missing. A person reviews the flags.
Two days → One hour
A sourced, structured answer to the question you actually asked, with the citations attached. A person reviews it and decides.
Two weeks → An afternoon
Write the brief and get back a built page: copy, imagery, layout, all on brand. Your team reviews it and hits publish.
Days of back-and-forth → The same meeting
Product shots, social tiles, banners, one-off artwork, generated to your brand guidelines. Pick the winners before the meeting ends, then post them.
AI use your team can see
One login, every leading model, full audit trail. The shadow AI already happening in your team moves somewhere you can see it.
Your handbooks, answerable
Load in your contracts, policies, and SOPs. Anyone can ask a question and get a cited answer in seconds, without interrupting someone senior.
Replace the tools you subscribe to
Build a client portal, a compliance tracker, or a CRM inside the platform, on your own data, and stop renewing the subscription it replaces.
Built for real work
Built into the platform.
Charts, approvals, agent rules, workflows and share analytics are included. Each answer keeps a record of how it was produced.
Figures, not paragraphs
Replies arrive as headline numbers, charts and tables, drawn from your own data.
Apps built from a description
An agent builds trackers, portals and dashboards beside the conversation, on your own data.
Rules you can read
Each agent shows what it does alone, what it asks you first and what it never does.
A yes before it goes out
Discounts, tenders and payments wait for a person, with the context to decide.
See what gets read
Share a report, an app or an agent and see how often it is opened. No cookies, no tracker.
Connectors
Works with the tools you already use.
Over 600 connectors, including these. Agents read and write through them with the permissions you set. See all connectors
- Gmail
- Google Drive
- Google Calendar
- Google Sheets
- Microsoft 365
- HubSpot
- Notion
- Jira
- Confluence
- Asana
- GitHub
- Dropbox
- Xero
- QuickBooks
- Shopify
- Stripe
Secure your data
Your data stays in your environment.
Documents, the search index, the audit trail and workflow logic stay inside your environment. Each model call can be set up as zero-retention, no-training and in-region.
Steps that only sort, label or check go to small classifiers rather than a large model, so less of your text is sent anywhere at all.
We can also set this up so you hold the contract with the model provider directly.
Your environment
- Documents
- Index
- Audit trail
- Workflow logic
- Agents and appsRun inside your instance
Governed model call
- Zero-retention
- In-region
- No training
Own your AI stack
Run it where you choose.
Start on the shared platform. Move to a dedicated instance when you need one, run by us or by your team, in our AWS account or yours. Everything you’ve built moves with it, and agents use only the models you approve.
It runs where you choose
Shared, dedicated in a cloud account we manage, or your own instance in your cloud account. Your documents, index, audit trail and workflow logic stay inside it.
Every output shows its working
Every field traced back to the document it came from, every validation recorded, every agent step reproducible. When a regulator, a client, or your own risk committee asks how a figure was reached, you answer from the record.
Already at work in regulated environments
Odokai is in use in regulated environments, where outputs go to external parties and are checked against compliance frameworks.
1 team
all sharing knowledge, no shadow AI
2+ hrs
saved on each case handled
100%
of outputs carry a provenance trail
Frameworks and harnesses are what developers use to build agents. Odokai is a governed workplace, already in use for regulated work, on your models, with the audit trail attached.
Reduce token costs
Match each step to the right model.
Each agent runs on the model that suits its job, from the providers you choose, and classifiers handle sorting and checking. The table shows what one million tokens costs on three common jobs.
| Example job | Tokens in / out | Top model for every step | Odokai, model matched to each step | Saving on model usage |
|---|---|---|---|---|
| Coding: reviewing and fixing codeSoftware and IT teams | 700k / 300k | $22.00 | $8.8430% top tier, 50% mid, 20% fast | 60% |
| Admin: inbox triage, invoices and updatesOperations and office teams | 800k / 200k | $18.00 | $0.8680% fast tier, 20% mid | 95% |
| Compliance: checking documents against policyLegal, financial services, insurance | 900k / 100k | $14.00 | $2.3260% fast tier, 30% mid, 10% top | 83% |
Illustrative, and model usage only: platform fees are separate. Providers' published standard API prices per million tokens, in US dollars, before caching or batch discounts, checked 20 October 2026: top tier $10 in / $50 out, mid tier $2 / $10, fast tier $0.10 / $0.50. Anthropic and OpenAI list their current three tiers at these prices; Anthropic's fast tier costs more on prompts over 100,000 tokens. Classifier steps are left out of the table. The model mix is set per workflow. On our cloud, model usage is part of your platform fee up to a fair-use volume; in your own AWS account you use your own model accounts.
- Support triage −89%
- Contract review −76%
- Board reporting −78%
- Tender drafting −85%
- Invoice matching −95%
- 38%ClassifiersRouting, labels, yes or no
- 33%Fast tierExtraction, summaries
- 21%Mid tierDrafting, analysis
- 8%Top tierHard reasoning
How it starts
Start with a discovery call.
Bring one job and a few of your own documents. In about 45 minutes we build an agent that does it, so you can judge whether it is worth automating. The call is free.
-
Step 1
See it work
Free ~45 minutes
Bring a handful of your own documents. On the call, we build an agent that handles one of your workflows, live, on your data. You leave with a working agent and a clear picture of what production-ready looks like for the rest.
-
Step 2
Production Rollout
Fixed fee 1–5 weeks
We take the agent you saw on the call and make it something you can put in front of clients: edge cases handled, plugged into your systems, every action logged, access controlled, your team trained on it. You sign off the acceptance criteria before we start and the fixed fee doesn't move after that. This is the only workflow that carries a full build fee. Once it's live, the platform and the governance underneath it are set up for good.
Scoped and quoted after the discovery call, once we both know what the workflow needs.
-
Step 3
Your Instance
Monthly fee shared or dedicated
What it costs to keep everything running once you're live. Model usage is included up to a fair-use volume we agree during your diagnostic, with a plain stated rate above it. Model upgrades reach your instance at no extra build cost. Your team builds new agents here without paying another rollout fee.
Talk to us
Talk to us.
Tell us about the job you want to automate. If it is a good fit, we arrange a 45-minute call and build the agent on your own documents.
Want to see it first? Browse the platform.