The three buckets of cost
Whenever someone asks "what does AI automation cost," the honest answer has three numbers:
1. Build — one-time, to design, build, integrate, and ship the thing. 2. Run — monthly API + tooling cost, varies with usage. 3. Manage — optional retainer to keep it tuned, monitored, and improving.
Marketing pages usually only show you the first one. Founders only think about the first two. The third is where most projects quietly die.
What builds actually cost in 2026
This is for small business work, not enterprise:
- Single focused automation: $2,500–$4,000. Lead follow-up, custom GPT, monthly reporting, document extraction. 2–4 week timeline.
- Connected set of automations: $5,000–$8,000. RAG chatbot + CRM integration, multi-step intake → quote → CRM, reporting + outreach pipeline. 4–6 weeks.
- Multi-team / multi-system rollout: $10,000–$25,000+. Internal copilots with SSO, agents with approval workflows, custom dashboards. 6–10 weeks.
Anything significantly cheaper is either a template you'll outgrow in a quarter or an offshore shop that won't be around when something breaks.
What it costs to run, monthly
The variable that surprises people. In 2026, with current model pricing, here's the honest range for a typical small-business automation portfolio:
| Volume | API spend | Tooling (n8n / Zapier / Make) | Total monthly |
|---|---|---|---|
| Light (1 automation, low volume) | $10–$40 | $0–$30 | $10–$70 |
| Moderate (3–5 automations) | $40–$150 | $30–$80 | $70–$230 |
| Heavy (chatbot + workflows + agents) | $150–$500 | $50–$200 | $200–$700 |
Versus the obvious comparison — even one extra hour of staff time per day is more than the heaviest end of that range.
What management actually costs
Here's the bucket nobody warns you about. AI automations don't stay shipped. Models get updated, deprecated, repriced. Prompts drift as the model changes. New edge cases come up. Integrations break when SaaS tools push updates.
You have three options:
1. Self-manage. Free, but only viable if you have someone technical and time-rich enough to actually do it. Most small businesses are not in this bucket. 2. Light retainer ($500–$1,000/mo). Monitoring, error alerts, a couple hours of changes a month, quarterly review. 3. Active retainer ($1,500–$3,000/mo). Everything above plus regular new automations, prompt tuning, monthly strategy. This is what most clients land on.
Skip this and the typical pattern is: project ships strong, runs great for 6–9 months, quietly degrades over the next 6, gets ripped out at month 18 because "AI didn't work."
We offer all three retainer tiers — Assist ($750), Operate ($1,500), and Scale ($3,000). The right answer depends on how much you're running and how mission-critical it is.
The ROI math we actually use
For each candidate automation, we ask three things:
1. Hours saved per month. Be conservative. If a task takes 30 minutes and happens 20 times a month, that's 10 hours. 2. Loaded cost of those hours. Use full burdened cost — salary plus everything else. For most small businesses that's $40–$80/hr. 3. Total cost of the automation. Build cost amortized over 24 months + monthly run cost + management share.
If (1 × 2) > (3 × 1.5), build it. The 1.5x buffer is for the inevitable surprises.
Worked example for instant lead follow-up:
- 5 hours/month of staff time saved replying to leads = 5 × $60 = $300/mo
- Increased close rate from sub-minute response: hard to quantify cleanly, so we leave it out of the base case
- Build: $3,000 ÷ 24 = $125/mo
- Run: $80/mo
- Management share: $200/mo
Cost: $405/mo. Conservative savings: $300/mo. Marginal.
Now factor in the close-rate lift (which is real but variable) — most clients see this go from "marginal" to "obvious" inside the first 90 days.
The mistake to avoid
Don't build something that saves 30 minutes a week. The math will never work and you'll resent the line item.
Pick the tasks that are eating real hours. Those are the ones with real ROI.
So what should you actually budget?
For a serious small business in 2026, planning to take AI seriously:
- Year one: $5,000–$15,000 in build cost, $1,200–$5,000 in run cost, $9,000–$36,000 in management.
- Year two onward: mostly run + management. Build cost only when you add new stuff.
That's not a small number. But it's also less than a single full-time hire, and the right portfolio of automations replaces 0.5–2 full-time hires worth of busywork.
If you want us to do the math on your specific use cases, book a free strategy call. We'll be honest about whether the numbers work or whether you should wait six months and start smaller.