What it costs

Pick the size of the problem.

The price depends on the number of systems, integrations and monthly volume. I confirm it in writing before work starts.

"From" is the starting price for the smallest scope of that sprint. Three things move it: the number of systems connected, the number of integrations and steps between them, and the monthly volume the workflow handles. Once you accept the written scope the price is fixed, and it changes only by written agreement.

Fixed price confirmed in writing. Build guarantee with a refund. No lock-in. You own what is built.

The three sprints compared

How these prices compare

"Typical final price" is my own estimate for a typical scope, not an observed average from client builds, and the written scope confirms the exact price before work starts. Other providers publish one-off build prices that I checked on 18 October 2026. They are marketing figures, not quotes, and their scopes differ from mine:

• Simple workflows $1,500 to $4,000 and medium builds $7,000 to $12,000 (Parix, updated 12 October 2026).
• Connected workflow automation $10,000 to $25,000 (SuperDupr, prices as of October 2026).
• Multi-workflow systems $5,000 to $15,000 (AgenticMarketingPro, updated 15 October 2026).

What most often raises a price: more systems connected, messy or duplicated data that needs cleaning first, and a higher monthly volume.

What it costs to run each month.

These are estimates, not quotes. You pay the vendors directly, in your own accounts. The figure that applies to you is set out in the written scope before work starts.

How the estimate is built

• Automation platform. Published prices checked on 17 October 2026: Make Core from $12 a month for 10,000 credits and Pro from $21 a month. Zapier Professional from $19.99 a month. n8n Cloud Starter at 20 euros a month (2,500 executions) and Pro at 50 euros a month (10,000 executions), both billed annually. Which one I use depends on your tools and volume.
• AI usage. Billed per use by the model provider. At about 3,000 tokens in and 500 out per item, one item costs about $0.011 on Claude Sonnet 5.5 ($2 and $10 per million tokens) and about $0.0006 on Claude Haiku 5.5 ($0.10 and $0.50 per million tokens). At 1,000 items a month that is about $11 or under $1. At 10,000 items it is about $110 or about $5.50. I choose the model per step, and keep cheaper models for simple steps.
• A worked example from my own site. The diagnostic report on this site is an AI workflow I run. Each report is capped at 1,000 output tokens. My server builds the prompt from fixed text and your five answers, and refuses anything over 3,500 characters, which is roughly 1,000 input tokens. On Claude Sonnet 5.5 at $2 and $10 per million tokens, that is at most about $0.012 per report in AI usage ($0.002 in, $0.010 out), so 1,000 reports a month would cost at most about $12. This is a ceiling from the configured limits and published prices, not a measured average.
• If your volume differs. The "Based on" row shows the volume each range assumes. AI usage rises in proportion to the number of items, and the platform cost steps up when you cross a tier limit. The written scope states your expected monthly volume and the tier it assumes. These ranges come from published vendor prices, not from measured client builds.
• A cost review inside your support window. Once the workflow has run for a couple of weeks, we look at your real platform and AI usage together, against the estimate in your written scope. If it runs higher than expected, I adjust the workflow, for example by moving simple steps to a cheaper model. This is part of the support or tuning window and costs nothing extra.
• Tools you already pay for. Your CRM, accounting, email and chat subscriptions are not counted. They are yours already.

These are calculated estimates from published vendor prices (Make, Zapier, n8n and the Anthropic API) as read on 17 October 2026, with headroom because each step in a workflow can use platform credits. I will replace them with observed averages from client builds once I have enough of them. If usage runs above the volume in your written scope, the vendor charges rise with it, and during the support window I flag any rise to you as soon as I see it. Vendors change prices and tiers, and higher volumes move you to a higher tier. I do not promise savings or hours saved.

A low price is not thin scope.

Each sprint is a fixed-scope build with the same working method, whatever its size. The price is lower because the scope is limited, not because the build is.

• Written acceptance criteria agreed before work starts, and the work is tested against them.
• A map of the workflow before anything is automated.
• Built in your own accounts, with credentials you grant and can revoke. You own what is built.
• Approval steps and written limits built into the automation.
• A handover: a recorded walkthrough and a one-page runbook.
• A support or tuning window of 30 days, or 90 days for the End-to-End AI Ops Engine.
• The build guarantee below, with a refund if the work still misses its criteria.

What you provide

• A signed data processing agreement before any data is shared.
• The tools named in the written agreement, in your own accounts.
• Credentials you grant and can revoke at any time.
• Timely access and approvals.
• Accurate source data.

How your data is handled, and the agreement I sign, are on the Trust page. I hold no security certification; the vendors I use publish their own.

I guarantee the build, not your results.

• Written acceptance criteria are agreed before work starts. They change only by written agreement.
• If the delivered work does not meet them, I fix it at no charge for up to 10 working days from your written notice.
• If it still does not meet them, I refund the fee you paid for that sprint. That fix or refund is your only remedy for the work missing its criteria.
• This does not cover problems caused by your changes, late access or approvals, incorrect data, or changes by third-party software or services.
• Tell me in writing, naming the criterion, within 14 days of handover. During the support window I also fix defects, meaning anything that does not meet an agreed criterion.
• I do not promise revenue, savings or hours saved. AI output can be wrong, so tests and approval steps are built in.

How it runs in time

1. Day 0: handover. The written acceptance criteria are what the work is tested against.
2. By day 14: you tell me in writing which criterion is missed. Notice given in this window keeps the refund remedy.
3. Up to 10 working days after your notice: I fix it at no charge.
4. If it still misses the criterion: I refund the fee you paid for that sprint.
5. Day 30 (day 90 for the end-to-end AI ops engine): the support or tuning window ends. Throughout it I also fix defects, meaning anything that does not meet an agreed criterion.

Questions about price and terms

Last updated: 18 October 2026