Capital and timeline

What it actually costs to reach a first paid pilot, and a grounded estimate of how long it takes

5 min read

Capital required

ItemMinimum-viableProperly-funded
Business entityAn existing entity reused, if its licensed activity already covers consulting; otherwise a freezone-style or equivalent light-touch local entity, typically a few thousand dollars all-in for the license plus any establishment/visa fees [Established, single-source cost figures — verify current fees directly with the registrar before committing]The same, plus a US entity (Delaware or Wyoming, formed via a remote-incorporation service, roughly $300–500 plus a modest annual registered-agent fee) for US-facing payment rails and contract perception — see Feasibility for a non-resident operator
LLM API (build and ops)Pay-as-you-go metered billing — a narrow single-workflow agent in pilot volume typically runs $20–150/mo [Established, verify at build time — provider pricing changes]A paid developer seat for build work, plus separate metered API billing for client-facing production
Orchestration (n8n/Make)A low-cost cloud starter tier ($24/mo), or self-hosted on a small VPS ($5–20/mo)A paid pro tier, roughly $50–150/mo per active client workspace
Vector database (for RAG builds)A free or low-cost managed tier — sufficient for most SMB-scale knowledge basesA dedicated managed vector database once client data volume or query load justifies it, roughly $70–300/mo
Voice AI (if offering a voice-agent product)A per-minute platform fee plus pass-through STT/LLM/TTS/telephony costs; a fully-loaded floor of roughly $0.08/min, realistic blended $0.10–0.30/min [Established, multi-source triangulated]The same stack, budgeted per client at roughly $150–400/mo for typical SMB call volume
Telephony numberA standard local number, roughly $1–2/mo plus usageThe same, multiple numbers for multi-location clients
HostingA free-to-low-cost tier$50–100/mo for client-facing uptime expectations
Portfolio-building$0 cash if built inside a business you already run — the highest-leverage optionA few thousand dollars if paying to build a demo case study for a vertical you have no existing foothold in
Legal (contracts, MSA, liability)A template MSA/SOW, a few hundred dollars one-timeA lawyer-drafted MSA with a liability cap, IP assignment, and data-handling clauses — a few thousand dollars, mandatory before any healthcare or financial-services engagement

Minimum-viable total to a first paid pilot: roughly $1,500–3,000 in cash, assuming an existing local entity is reusable and build tooling stays in pay-as-you-go tiers — genuinely low, because AI-tooling fluency and an existing operating stack absorb most of the technical capital requirement. Properly-funded (a dedicated US entity, lawyer-reviewed contracts, a paid-for demo build in a new vertical): roughly $8,000–15,000.

Realistic timeline to first dollar

This is a dependency chain, not a calendar — each step gates the next.

  1. Niche selection — 3–5 days. Finalize, don't re-litigate. The lowest-friction entries are a niche adjacent to a business you already run, or a niche your existing client base already clusters in — property management, legal, healthcare-adjacent local services, and similar are worth auditing from an existing client list before looking externally.
  2. Proof-asset build — 1–4 weeks. Not a generic demo — a working narrow automation, built either inside your own operations or offered free or heavily discounted to one existing client in the target vertical, in exchange for a case study and testimonial. The clearest documented "first client" pattern found in operator accounts is a founder-owned distribution channel used as a live testing ground before any external sales motion — reducing a multi-week manual process to a few hours became the sellable proof point in the sourced example. [Directional — a single self-reported Indie Hackers account, not independently verified, dated, or backed by disclosed figures a third party could check; it illustrates a plausible and mechanically sensible pattern, not a confirmed result.] An owned distribution channel (your own business, your own client base) is a structural advantage most first-time AI consultants don't have, and should be used before any cold outreach.
  3. Outreach to a first external client — 2–4 weeks, in parallel with step 2. Warm channel first — existing clients, where trust and billing infrastructure already exist — then cold or LinkedIn outreach only once a proof asset exists, since outreach converts far better against a demonstrable artifact than a claim.
  4. First paid pilot close — 1–2 weeks of sales cycle once a real prospect is engaged with a concrete proof asset in hand. Sales cycles for a narrowly-scoped $5–15K pilot are short if the buyer can see it working; they stall indefinitely on abstract "AI transformation" pitches.
  5. Delivery — 2–4 weeks for the narrow pilot.
  6. Case study — 3–5 days to package, once the pilot has been live for 2–3 weeks and there's real usage data, not day-one enthusiasm.
  7. Expansion sale to the same client, or the case study closes client #2 — ongoing.

Grounded estimate: first dollar in 4–8 weeks, if you use an owned channel (your own client base) for the first sale rather than cold outreach, because steps 2 and 3 collapse into the same relationship. Discount "get your first client in 7 days" content aggressively — the evidence found across operator accounts and trade press shows real first engagements come from pre-existing distribution far more reliably than cold outbound, and fast-cold-outreach claims are uniformly course-seller marketing with no disclosed operator case behind them. [Established pattern from available evidence; the "fast cold outreach" claims themselves are speculative/discredited.]

AI Agency · progress saved in this browser · sign in to sync across devices

Up next

Feasibility for a non-resident operator

The payment-rail friction is real, pre-solved by anyone who's already sold to US customers, and this business has no licensing wall to work around in the first place

3 min