The side-by-side
| Build it yourself | Agent Lease | |
|---|---|---|
| Initial build | ~2–6 weeks engineering time per agent | ~5–10 days, done for you |
| Tooling cost | API tokens + framework subscriptions, £100–£500/month | Included in £6/day |
| Engineer time | Yours, or a contractor at £500–£1,200/day | None |
| Maintenance | Ongoing - model updates, broken integrations, prompt tuning | Included |
| Model upgrades | You re-test, re-tune, redeploy each time | We do it, no action from you |
| Watchdog & safety | You design + build it | Included |
| Integration with your tools | You write the connectors | Included |
| WhatsApp brief layer | You build it | Included |
| Cancellation | You own the code, but maintenance stops if you stop | Month-to-month |
The cost line nobody draws
The build cost is often quoted as just “engineering days to ship v1.” The real cost is in the ongoing maintenance. Three things eat ongoing time:
- Model churn. Frontier AI models release new versions every 2–4 months. Each one needs your prompts re-tested, your agent re-tuned, your evaluations re-run. A small business that built on GPT-4 in 2024 has had to re-do this 3–4 times by 2026.
- Integration drift. Email APIs change, calendar permissions tighten, your CRM does an upgrade. Each one is a 1–3 day fix. Skip a few and the agent quietly degrades.
- Voice drift. Your agent’s tone wanders over time as the model updates. Keeping it on-voice is an ongoing job, not a one-time setup.
A realistic maintenance load for one DIY agent is 1–2 days of engineering time per month. At contractor rates, that’s £500–£2,400 per agent per month. The build cost is the down-payment; the maintenance is the rent.
Where building yourself genuinely wins
- You have an engineer on staff who’d enjoy the work. If maintenance lands on someone who’d find it interesting, the cost is real but the energy isn’t.
- Your needs are highly bespoke. If your business requires deep integration with a niche tool nobody else uses, you may need to build it. We’d be candid if your case is this.
- You see the agent itself as IP. Some businesses want to own the agent for strategic reasons (resale, productisation). Owning your code matters here in a way it doesn’t for most owner-operators.
Where the lease wins
For most owner-operated businesses, three things tip the balance to lease:
- Your time is the constraint, not your money. A £180/month lease vs £500–£2,400/month in maintenance is a real saving, but the bigger one is that the maintenance isn’t on your plate. You get to work on the business, not on the agents.
- You don’t want a new dependency. If your engineer leaves, your DIY agents stop being maintained. A lease keeps running.
- You can cancel. If it turns out an agent doesn’t fit your business, you stop the lease. You can’t un-spend the engineering time on a DIY agent.
How to decide for your business
The simplest test: if you have an engineer who would want to own the agent for the next 18 months, build. If you don’t, lease. The break-even on cost alone (lease vs build with a contractor) usually lands inside 3–6 months in favour of leasing.