Most AI automation agencies do not publish pricing. The buyer guides keep saying it. The reviews keep saying it. Open ten agency websites today and you will see the same pattern: feature lists, case study teasers, and a "Contact us" button that opens a discovery flow before anyone says a number.
This post publishes the numbers. All of them. Sourced, real, and reproducible from public material so you can sanity-check our retainer against the rest of the market before you book a single call.
The five line items every AI agent engagement has
Across audit, pilot, build, operate, and recovery work, every legitimate engagement charges for five things. Knowing which line item is fixed-fee vs variable is the first step in pricing your own buying decision.
1. Audit or readiness assessment
The market floor for a 2-to-4-week AI readiness audit is around $5,000. The ceiling for the same scope tops out around $15,000 when an enterprise consultancy attaches a partner-level interview process and a 30-page report. Industry data: DigitalAgencyNetwork's 2026 pricing guide and the CFO-focused breakdown by Optimize with Sanwal both anchor on this range.
What you should get for $5,000 to $15,000: a written deployment plan, an inventory of which workflows are agent-ready vs operator-ready vs neither, a stack recommendation, and an honest read on whether the engagement is worth pursuing at all. If the audit deliverable is a slide deck with no specific runbook recommendations, you paid too much.
DPL prices the audit at $0 because we recover the cost inside the first month of any engagement that turns into a pilot. The trade-off is full transparency: the audit goes through a co-founder, not a junior consultant, and we tell you when we cannot help.
2. Pilot. One workflow, scoped and shipped
Pilot pricing splits into two market segments.
For a simple workflow automation (one trigger, two-to-three steps, one API connector), $1,500 to $2,500 fixed-fee is the going rate. The DPL pilot lands at $2,500 fixed for 30 days, scoped to one workflow operated in production with a real-time Slack channel attached. The pilot is operationally loaded. Meaning we run the workflow during the 30 days, not hand you the build and walk away.
For a more complex multi-step AI agent build (LLM planner, three or more tool calls, state persistence, retry logic, an operator review queue), $5,000 to $15,000 fixed-fee is typical. Arsum's 2026 pricing breakdown and Braincuber's pricing guide both place the bulk of pilots in this band.
What the price floor of $1,500 buys you in 2026: a Zapier-only build with no observability layer, no retry strategy, and no operator on call. That is a flag, not a feature. Buyer guides explicitly say demand the tech stack upfront and treat Zapier-only as a yellow flag.
3. Custom development
If the engagement scope is "build me a real production agent from scratch," the market range jumps to $25,000 for a structured MVP and $300,000+ for a true enterprise-grade agentic system. Custom voice agents (branded customer service bots, AI narrators) run $2,000 to $25,000+ depending on the depth of integration and the number of intent paths.
Salesforce Agentforce implementations sit at the top of this range. $50,000 to $150,000 implementation plus $10,000 to $25,000 per month ongoing consulting per the Fin.ai pricing comparison. This is where enterprise AI buyers end up if they shop on the "established platforms" axis instead of the agency-managed axis.
4. Monthly retainer. Production operations
Once the agent is shipping production work, the engagement shifts to retainer.
| Tier | Monthly retainer | What it buys |
|---|---|---|
| SMB / off-the-shelf | $500 to $5,000 | Monitoring, model swaps, retry queue management. One or two workflows. |
| Production-grade | $3,000 to $15,000 | Multiple production workflows, drift detection, weekly narrative reports, operator escalation. |
| Enterprise | $10,000 to $25,000+ | Compliance-sensitive workflows, custom integrations, on-call SLAs. |
DPL sits in the SMB tier at $2,500/month, deliberately. The math anchor that makes it work: a four-person ops team (Ops Manager + Reporting Analyst + Lead Coordinator + QA Reviewer) at $80K-$100K base each is approximately $400K fully loaded annually. Our retainer at $30K/year is 13× lower. The savings exist because the AI runs the workflow, operators audit exceptions, and co-founders sign off. There is no 6-person account ladder to fund.
5. The hidden 20-40% integration tax
The single most-underbudgeted line item is integration. The Crunch's AI agents price guide puts the integration overhead at 20% to 40% on top of the platform or retainer fee. This covers connecting your CRM, helpdesk, billing, identity provider, and any internal data sources to the agent. It also covers data preparation and the operator-hours spent training the agent on your specific edge cases during the first month.
Token costs are the second line item people miss. Most production agents using Anthropic Claude or OpenAI GPT-4-class models will spend $50 to $500 per workflow per month at moderate volume (10-50 invocations per day). High-throughput workflows that hit the LLM thousands of times daily can spike to $2,000+ per month per workflow.
What this means for buyers in three sentences
If the agency lists pricing publicly, anchor your evaluation on whether their math story holds. What they charge versus what the in-house alternative would cost.
If the agency refuses to publish a range and only offers a discovery call, treat that as a 20-40% price premium baked into the obscurity.
If the agency leads with platform-licensing arguments and the retainer is mostly seat fees and not operator hours, you are buying a Notion of contractor work dressed up as automation, not a production agent stack.
The DPL pricing card
We publish ours in full and stand behind it on the /pricing page.
- Audit: Free. 45 minutes with a co-founder. Written deployment plan within 5 business days.
- Pilot: $2,500 fixed for 30 days. One workflow scoped, built, operated.
- Retainer: $2,500/month. Production operation, replacements as we ship them, monthly written narrative report. Cancel any month.
- Recovery Diagnosis: $5,000 fixed for 2 weeks. 30-criteria audit of an existing agent stack. Written report and 1-hour walkthrough call.
- Recovery Fix: $10,000 fixed for 4 weeks. Production fix with proper observability, retries, rollback path, and operator handover.
If you want to vet our pricing against the rest of the market before booking, run our 10-question diagnostic. It scores your existing agent stack against the same 30-criteria checklist that drives our recovery service. No email required.