Grounded in your data.
Every AI feature we build works from your real Odoo data. If the answer isn't in your system, it won't make one up.
AI @ ADVANCE INSIGHT
AI THAT WORKS.
We use AI every day, and we build AI features our clients actually use. It writes step-by-step guides with screenshots, mocks up new screens, and turns raw numbers into reports you can read. We fix bugs in a fraction of the usual time by letting AI take the first pass. This is just the start, and every week we put AI to work on more everyday jobs: sales, invoicing, customer follow-up.
portal_description_generator
input task 2184 · 18 timesheets · 6 comments · 3 stage changes
model claude-sonnet-4-6 via ai.composer
... gathering context
... drafting summary
ok 194 words · posted to portal · ready for client
How we think about it
AI where it earns its keep.
We don't sell "AI transformation". We build small AI tools that earn their place in a real African business, and we only put one to work when it genuinely saves time.
Humans stay in the loop.
AI writes the first draft. A real consultant checks it and approves it before anything reaches a client, a supplier, or your accounts. AI is an extra pair of hands. It never has the final say.
You're not locked in.
We can use ChatGPT, Claude, or Odoo's own built-in AI, and switch between them without rebuilding anything. You're never tied to one provider, and your data stays yours.
What we've put to work
Five AI tools we use every day.
Each one started as a job somebody was spending too long on. AI does the slow first pass, and a consultant makes sure it's right.
Plain-English updates on every project.
Log in to our client portal and every project and support ticket has a short, clear summary of where things stand. AI reads everything behind the scenes, the notes, the history, the hours logged, and writes a clear update of what happened and what's next. One of our team checks it before it's published. You get a straight answer without chasing anyone, and we save the half hour it used to take to write by hand.
- Built from the real project history, notes and hours, not guesswork
- Works with Claude, ChatGPT or a private model, your choice
- Checked by a real person before you ever see it
BEFORE
stage: In Progress → Client Review
timesheets: 6.5h (A. Wanjiru, D. Kiprop)
chat: "STK push callback fails on 60s+"
chat: "fixed, retrying, see PR #48"
AFTER
"We traced the paybill timeout to STK push
callbacks dropping after 60 seconds. The fix
is in staging and handed to your team for
UAT. We'll promote on sign-off."
CONSULTANT TAPS
AGENT DRAFT
"Manufacturing runs across two plants with
subcontracting to a third-party assembler.
The client reports through eTIMS in KES. MRP,
work orders and BoM routing are in scope;
quality checks and maintenance follow in
phase two."
Your project plan, written while we talk.
When we sit down to plan your project, we capture the conversation as it happens, a tap here, a note there. As we finish each part, AI turns that shorthand into a proper written plan, flags anything we forgot to ask, and drops the phases, tasks and time estimates straight into your project file. What used to be a two-day write-up is ready before the meeting is over.
- Captures the conversation live, as we talk
- AI turns shorthand into a full plan and spots the gaps
- Phases, tasks and hours land straight in your Odoo project
- One clear plan everyone works from
An MCP server that lets Claude drive Odoo.
We run Model Context Protocol (MCP) in production. Our MCP server for Odoo lets Claude, or any AI assistant that speaks MCP, search records, read fields, create and update documents, and call model methods with proper permission checks. Paired with a companion Odoo module, it gates every call through an API key, a per-model allowlist, record rules and an audit log. Our own team uses it daily; it's the plumbing behind most of the other builds on this page.
- Full CRUD over any Odoo model, with granular field selection
- API-key auth, per-model allowlist, record rules enforced
- Tested against Odoo 17, 18 and 19 (CE and Enterprise)
- Published under MPL-2.0: use it on your own instance
tool: odoo.search_records
model: sale.order
domain: [("state", "=", "draft"),
("partner_id.country_id.code", "=", "KE")]
fields: [name, partner_id, amount_total]
→ 14 orders returned
/scope-create-phases
phase plan from shorthand
/scope-create-tasks
tasks + hour estimates
/odoo-code-review
custom module audit
/pre-golive-test
XML-RPC regression suite
/monthly-pl
branded P&L from Odoo
A toolbox that does the busywork.
Every consultant on our team has a set of AI helpers for the repetitive parts of the job: turning interview notes into a project plan, checking our own code for mistakes, testing a system before it goes live, and producing the monthly profit-and-loss report. The AI does the first pass in seconds; the consultant still owns the result. Work that used to take hours now starts from a solid draft.
- Interview notes turned into a full project plan
- Our code checked for mistakes before it goes live
- Systems tested before they go live
- Monthly profit-and-loss reports straight from your data
Slack that knows what Odoo knows.
Our team works in Slack and Odoo all day, so we connected the two. Reminders to log hours, a daily round-up of what's moving, a heads-up when a sales order is ready, even an out-of-office status that sets itself from people's leave, all flow from Odoo straight into Slack. AI sums up the noise: "Six tasks moved forward today, here's where." People read one short message instead of digging through the system. We can set the same thing up for clients on their own channels.
- Daily reminders and updates posted automatically
- Out-of-office status that updates itself from leave
- Plain-English daily summaries instead of raw noise
- We can set the same up on your Slack
#delivery-daily
08:30
Yesterday across 12 projects:
- 6 tasks moved to Client Review
- 2 estimates approved (KES 1.4M)
- 1 go-live scheduled for Friday
- 3 tickets waiting on client input > 48h
Have an AI idea for Odoo?
Tell us which job is eating your time, and we'll tell you honestly whether AI is worth it, and what a first version would look like.