Azimut AI
Illustrative scenario. The company on this page doesn't exist: it's a typical profile for the sector, built to explain how Azimut would be used. It doesn't describe any customer or claim measured results.
The company in this scenario
- 250 people: 190 engineers across energy, infrastructure and industry, and 60 in support functions.
- Project documentation in SharePoint, after migrating the old file server; email and Teams.
- Bids on eight to ten tenders a month, with specifications of 60 to 200 pages.
- Several senior project managers a few years from retirement.
The problem
An engineering firm's edge is its accumulated experience, and it lives in closed projects nobody opens again. Every technical bid starts by rewriting the method statement, work plan and references; every tender gets read in full hunting for the requirement that disqualifies you. Lessons learned get written when a project closes and the next one never reads them. And when a senior project manager retires, a good part of their area's technical judgment goes with them.
How it would work with Azimut
Azimut would index the documentation repositories with the permissions they already have, so each engineer only sees the projects they can already access. The firm would publish four agents.
| Agent | What it does | Who uses it | What gets measured |
|---|---|---|---|
| Technical bids | Writes the method statement, methodology and work plan from the tender, reusing winning bids and CVs. | Bid team | Hours per bid and award rate. |
| Tender review | Extracts requirements, scoring criteria and grounds for exclusion, and checks the bid against them before submission. | Bid team and project managers | Bids excluded on a formal defect. |
| Project lookup | Answers "how did we solve this on another project?" with past project documentation, citing the source document. | Engineers | Time to find the precedent. |
| Lessons learned | Captures what was concluded when each project closed and surfaces it when a new project looks similar. | Project managers | Lessons reused per project. |
Bids and tender review call for the top-capability model: long documents where one detail changes the outcome. Technical lookups work well with a mid-tier model. Engineers pick the model per task from OpenAI, Anthropic, Google and Mistral.
Where you'd start
You'd start with tender review. Its value is immediate and easy to check, because a bid excluded on a formal defect is all that work lost; it's used on every tender and doesn't require changing how bids are written. Meanwhile, the repositories the project lookup agent will need get connected and organized.
What's out of scope
Azimut works with documents (method statements, reports, specifications, minutes), not with CAD or BIM models or calculation software: there's no connector for those tools and it doesn't read drawings the way an engineer would. The agent doesn't sign off or validate a calculation; technical judgment and responsibility remain the engineer's. Integrations are confirmed during rollout and only the tested ones get activated.
How it would be measured
Before turning on any agent, you set the baseline, with the company's own numbers rather than industry averages.
Week 0. Hours per bid over the last ten, award rate over the last year, bids excluded on formal defects, and current AI use in the firm: who, with which tool and at what cost.
Month 3. The same numbers, next to Azimut's usage panel by user, area, model and tool.
Frequently asked questions
Does it read drawings?
Not the way an engineer does. It works on the text of method statements, reports, tenders and specifications. CAD and BIM models are out of scope.
What about project documentation under confidentiality clauses?
Agents inherit source permissions: if a folder is restricted to a project team, only that team gets answers from it.
Do we need to clean up the whole archive first?
There's no need to start there. You connect the bid repositories and the most recent projects first, and expand as it gets used.