How we work
Engagement team
AI projects fail when the people who sold the work disappear and juniors learn on your dime. Our engagement model assigns clear roles with explicit accountability. The engagement lead stays from Perceive through Deliver; architects and tech leads rotate in as depth requires; Run has dedicated operations engineers who know your system history.
Discovery team — three people, two weeks
Perceive runs with a fixed trio: engagement lead, solution architect, data engineer. The engagement lead owns stakeholder alignment, commercial clarity and readout quality. The solution architect owns target architecture, integration approach and technical risk. The data engineer owns data source verification, readiness assessment and pipeline feasibility.
All three are senior enough to commit the firm to the fixed-price proposal they co-author. Junior staff may assist with research; they do not lead workshops or sign the pack.
Build squad — scaled to scope
Engineer and Deliver deploy a squad sized to the fixed scope. A typical first production release squad includes: engagement lead (part-time), tech lead, two to four product engineers, a data or ML engineer, and UX design support where the use case has a user interface. Security review draws from a pooled specialist; evaluation design may add a dedicated engineer for complex agent flows.
- Engagement lead: backlog prioritisation, steering prep, scope change negotiation.
- Tech lead: architecture integrity, code review standards, release readiness.
- Product engineers: agent logic, integrations, API layers, front-end where needed.
- Data / ML engineer: retrieval, evaluation harness, model integration, performance.
- UX: task flows, error states, operator dashboards when in scope.
Squads are cross-functional within the sprint. We do not throw requirements over a wall to a offshore factory. Our engineers sit in Bengaluru, Pune and the USA; distribution follows your timezone and data residency needs, not a race to the cheapest hour.
How we embed
Embedding depth is agreed in discovery. Options range from fully remote with structured ceremonies to on-site weeks during integration spikes. We match your engineering culture where possible: your Jira or Linear, your Slack or Teams, your git platform. We add our evaluation and observability tooling where yours lacks AI-specific coverage.
Pairing with your developers is encouraged when you want internal capability. We document as we build so your team can maintain without us. Run is easier to exit when knowledge transfer was real.
Governance rhythm
- Daily: async standup in shared channel.
- Weekly: sprint demo and retrospective with product owner.
- Fortnightly: steering with executive sponsor — risks, decisions, budget.
- At Deliver: operational handover to your team or Run.
- Under Run: monthly operations review, quarterly roadmap.
Escalation is direct. If blockers sit with your side — access, decisions, third parties — the engagement lead names them in steering with dates, not vague delays.
Run team
Run is staffed by engineers who operate production systems, not a generic service desk reading scripts. Typical coverage: primary on-call engineer, backup, engagement lead for commercial and roadmap conversations, security liaison for incident classes that touch policy. Coverage hours depend on contract; business-hours plus on-call for critical systems is common for the first operational year.
What we ask from you
- A named executive sponsor with authority to decide scope trade-offs.
- A product owner available for weekly backlog and demo sessions.
- Timely access to systems, people and environments — delays consume calendar, not just effort.
- Security and compliance counterparts when regulated data is in scope.
- Honest feedback when demos miss the mark — early correction beats late rework.
The method works when both sides hold up their side. We are explicit about client dependencies in the proposal so week twelve remains credible.
Seniority and continuity
The people in discovery are senior enough to commit the firm to price and date. The tech lead in build is senior enough to reject shortcuts that would fail production review. Run engineers are senior enough to diagnose model regressions without escalating every alert to the original build team. That depth costs more per hour than a body shop; it costs less than a failed go-live.
When team members rotate — holiday, illness, growth — overlap is built in. No single person holds irreplaceable tribal knowledge without documentation backup. Your engagement lead remains the constant commercial and delivery contact across phases.
Will the same architect stay through build?
The architect who wrote the discovery architecture is involved through Engineer at minimum, often through Deliver. Continuity is deliberate.
Can you work entirely on-site?
Yes for defined periods when co-location helps — workshops, cutover, training. Permanent on-site for the full build is priced explicitly if required.
Do you subcontract niche work?
Rarely, and only with your approval. Core engineering stays in-house to Incubics. Specialist penetration testing or niche hardware may use approved partners under NDA.
Next step
Start with two weeks.
A fixed-fee discovery gives you a ranked use-case portfolio, a target architecture, a cost model and a build proposal you can take to your board. If we don't find a case worth building, we tell you.