Custom AI/Timeline
How long does custom AI software development take?
Most projects ship in 3 to 12 weeks. Here’s the week-by-week breakdown of what I do, what you get at each step, and what speeds it up or slows it down.
The four phases
Every custom AI project, from a focused chatbot to an enterprise rollout, runs through these four phases. Only the depth changes.
Week 1–2
Discovery & scope
We map your workflow, agree the scope and write the technical spec. You sign it off before any code is written.
- Free 30-min call + written scope
- Process diagram of where AI fits
- Fixed-price quote with what's in / out
- Data access checklist
Week 2–8
Build & integrate
The bulk of the project. We build the AI logic — agents, tools, memory, prompts — and wire it to your systems. You see weekly demos.
- Working prototype by end of week 2
- Live demo every Friday
- Each integration tested in isolation
- Audit log + evaluation harness
Week 1–3
Test & harden
Edge cases, security review, user acceptance testing. Real users run real tasks against the system before launch.
- Edge-case test plan + run
- Security review (data flow, secrets, PII)
- Admin panel + permissions tested
- User acceptance sign-off
Week 1–2
Launch & support
Deployment to your server or our managed cloud. 30 days of post-launch support included — bug fixes, tweaks, training.
- Production deployment
- Team training session (recorded)
- 30 days included support
- Ongoing managed plan available
What makes projects faster or slower
Same project, different conditions, very different timelines. The speed factors below explain most of the variance.
- Clean, well-documented APIs on every system
- Single decision-maker on your side
- Existing AI patterns we've shipped before
- Data already exists in a queryable form
- Managed cloud deployment
- Legacy systems with no API or undocumented data formats
- Multiple stakeholders + committee approval at every step
- Novel use case that needs research and prototyping
- Data trapped in PDFs, scanned docs or spreadsheets
- On-premise + custom security review process
Real project timelines
Anonymised examples from past work. Use them as a gut check against your own scope.
| Project | Integrations | Timeline |
|---|---|---|
| WhatsApp customer support bot | 2 systems | 4 weeks |
| Bank reconciliation + invoice AI | 4 systems | 8 weeks |
| Internal AI assistant for legal team | 3 systems (DMS heavy) | 10 weeks |
| Multi-department AI ops platform | 12+ systems | 16 weeks |
Timeline FAQ
How long does it take to build a custom AI?
A focused project — chatbot, single automation — ships in 3–6 weeks. A typical multi-integration build runs 6–10 weeks. Enterprise rollouts with on-premise hosting and compliance review take 12–20 weeks. The pricing page narrows the timeline down by project type.
Can you do it faster?
Sometimes. If your need fits an existing ClawdBot template, you can be live in days on a managed plan. The 3+ week minimum applies to genuinely custom builds where we're writing new logic, integrations and agents.
What slows projects down most?
Data access and decision-making. We've had projects sit idle for 3 weeks waiting for SSO credentials or a CSV export. We flag access dependencies in the scope doc on day one to minimise surprises.
When can we start?
I personally take on 2–3 projects at a time. Current lead time from sign-off to kick-off is usually 1–2 weeks. Drop me a message and I'll tell you the actual next available slot.
Do you offer phased rollouts?
Yes — and for larger multi-system projects I usually recommend it. Phase 1 ships a working slice in 4–6 weeks so you start getting value, then we layer in the rest over the following months. Better than a 6-month silent build.
Ready to start the clock?
Free 30-minute discovery call. By the end you’ll know the scope, the timeline, and a written fixed-price quote within a few days.
Mark Austen, Founder — replies within 24 hours