TLDR
- This post is about the project cost, not the license cost. For Flex Credits, per-conversation, and per-user rates, see the Agentforce pricing guide.
- Services typically run two to five times first-year platform spend. Business cases built on license price alone are wrong by that multiple.
- Data remediation is the single most variable line and the one most often discovered after signature.
- Partner day rates span roughly $500 to $3,500 depending on delivery model. Rate matters less than scope discipline.
- Ongoing admin capacity is a permanent operating cost. Budget it or it comes out of your existing team's slack, which is usually zero.
- Every figure here is a typical 2026 market rate for Salesforce implementation and consulting work, not a quote. Your actual cost moves with org complexity, data quality, and scope.
For the license and Flex Credit breakdown, including per-action, per-conversation, and per-user rates, see Agentforce pricing explained. That page covers what Salesforce charges you. This page covers everything that price does not include, which is where most of the budget variance lives.
Why the License Number Misleads
Agentforce pricing is unusually transparent for enterprise software. You can calculate expected platform spend from conversation volume and action counts with reasonable accuracy, and the pricing guide linked above does exactly that.
That transparency creates a specific failure. A finance team gets a credible platform number, approves against it, and then meets the implementation quote in month two. At that point the project either stalls, or it proceeds with the services work compressed to fit the leftover budget. Compressed services work means skipped remediation and skipped testing, which produces the pilot failures described in why Agentforce pilots fail.
The platform line is a utility bill. The services line is a construction project. They behave differently and they should be estimated separately.
The Six Services Cost Lines
1. Discovery and scoping
Typical range: $8,000 to $30,000.
Use case definition, current-state assessment, data quality baseline, automation inventory, permission model review, success metrics, and a delivery plan. Two to four weeks.
Teams try to skip this because it feels like paying for a proposal. It is the line with the best return in the whole project, because it is what converts a fixed-price build into an accurate one. Skipping discovery does not remove the work, it moves it into the build phase at a worse rate and without a change-control conversation.
If your partner offers discovery for free, understand what that means: the cost is inside the build price, and their incentive is to reach the build quickly rather than to tell you the project should be smaller.
2. Data remediation
Typical range: $15,000 to $150,000+. The widest line by far.
Duplicate resolution with survivorship rules, ownership and territory correction, picklist consolidation, field population backfill, dead field deprecation, and fixing the intake processes that caused the problems.
The range is enormous because it depends on things nobody knows before assessment: how many years of configuration have accumulated, how many admins have owned the org, whether there was ever a data governance function, and how many integrations write to the same objects.
Rough calibration from assessments:
| Org condition | Typical remediation effort |
|---|---|
| Recently implemented, governed, single admin owner | 1–2 weeks |
| Mature, documented, some drift | 3–5 weeks |
| Mature, multiple admin generations, no governance | 6–10 weeks |
| Post-acquisition, merged orgs, or heavy custom sprawl | 10–20 weeks |
This is also the line where an AI agent changes the economics. The work is high-volume and repetitive: identify duplicates across a large record set, propose merges, flag ownership mismatches, find fields populated on 4% of records. An agent working through that backlog with an admin approving batches covers ground that a human team would need months to reach. It does not remove the admin, it removes the reason the backlog never got started.
3. Data 360 build, only if you need it
Typical range: $0 to $200,000+.
The zero is real and it applies more often than vendors suggest. Agents grounded on CRM records through Flow, Apex, and prompt templates do not need a Data Cloud modelling project. The full argument is in does Agentforce require Data Cloud, and it is worth reading before you accept a quote that includes this line.
When it is genuinely required, the work is source analysis, data stream configuration, data model mapping, identity resolution rulesets, retriever and index configuration, relevance tuning, and validation. Cost scales with source count and with how badly the sources disagree about identity.
"Which grounding path did you evaluate, and why was prompt template and Flow grounding ruled out for this use case?" A partner who cannot answer specifically has scoped the platform build by default rather than by analysis. That single question is often worth six figures.
4. Agent build and integration
Typical range: $20,000 to $120,000 per agent, declining sharply after the first.
Topic and instruction design, action development, prompt engineering, integration with external systems where needed, and iteration against real usage.
The first agent carries the platform learning curve, the pattern decisions, and the integration groundwork. The second agent in the same org, reusing those patterns, commonly lands at 40 to 60% of the first agent's cost. This is why "we'll start with five use cases" is worse value than it sounds, and why sequencing matters more than scope breadth.
5. Testing and governance
Typical range: $12,000 to $60,000.
Scenario library construction, persona-based permission testing, action testing including malformed input, cascade testing against existing automation, audit trail implementation, rollback tooling, approval workflow design, and consumption monitoring.
This is the most frequently cut line and the one whose absence causes the most expensive failures. An agent without a tested rollback path is an agent your risk committee will eventually stop. Cutting this line does not save money, it defers it into incident response at a much worse rate.
6. Enablement and handover
Typical range: $6,000 to $30,000.
Admin training, documentation, runbooks, and the structured handover that lets your team own the agent after the partner leaves.
Cut this and you have bought a dependency. Every prompt tweak becomes a support ticket at day rates. Over three years the retainer exceeds what proper handover would have cost several times over.
Partner Day Rates in 2026
Market ranges for Salesforce and Agentforce delivery. These are rates, not quotes, and the delivery model matters more than the headline number.
| Delivery model | Blended day rate | Best fit |
|---|---|---|
| Offshore delivery team | $400 – $800 | Well-specified, high-volume remediation work |
| Nearshore / hybrid pod | $700 – $1,300 | Mixed build and remediation with timezone overlap |
| Onshore boutique / specialist | $1,200 – $2,200 | Complex orgs, architecture decisions, regulated environments |
| Large systems integrator | $2,000 – $3,500 | Multi-region enterprise programmes with formal governance |
| Salesforce Professional Services | $2,200 – $3,500 | Platform-edge cases and escalation paths |
How to read this. A cheap day rate on a badly scoped project costs more than an expensive rate on a tight one, because rework is billed at the same rate as work. The variable to control is days, and days are controlled by scoping quality.
The pattern that works for most mid-market orgs is a hybrid: senior onshore capacity for architecture, data model decisions, and governance, with lower-rate delivery capacity for the high-volume remediation execution. Paying $2,000 a day for duplicate merging is waste. Paying $500 a day for sharing model architecture is a different and worse kind of waste.
For general Salesforce consulting rate context beyond Agentforce, see Salesforce consultant cost.
Pilot to Production: Where the Time Goes
The build is not the long pole. This is the timeline shape we see for a first agent in a mature org.
| Phase | Duration | Services cost share |
|---|---|---|
| Discovery and assessment | 2–4 weeks | 10–15% |
| Data remediation | 3–10 weeks | 30–45% |
| Agent build | 3–6 weeks | 20–30% |
| Testing and governance | 2–4 weeks | 10–20% |
| Controlled pilot | 6–8 weeks | 5–10% |
| Production rollout and enablement | 2–4 weeks | 5–10% |
Total elapsed time is typically 16 to 30 weeks for a first production agent, with phases overlapping. Remediation and build can run partly in parallel once the assessment is complete.
Note that remediation is the largest single share. Any quote where remediation is a small fraction of total cost is either working in an unusually clean org or has not assessed the org yet. Ask which.
First-Year Total Cost Scenarios
Services only. Platform and license spend sits on top, and those numbers come from the pricing guide.
Scenario A: Single-use-case pilot
Profile: 50 to 200 Salesforce users, one internal agent, CRM-record grounding, no Data 360, reasonably maintained org.
| Line | Cost |
|---|---|
| Discovery and scoping | $8,000 – $15,000 |
| Data remediation | $12,000 – $30,000 |
| Data 360 build | $0 |
| Agent build | $15,000 – $25,000 |
| Testing and governance | $8,000 – $15,000 |
| Enablement | $5,000 – $10,000 |
| First-year services total | $48,000 – $95,000 |
Plus roughly 0.25 FTE of ongoing admin capacity.
Scenario B: Mid-market production rollout
Profile: 200 to 1,000 users, two to three agents, one requiring knowledge grounding, several years of accumulated configuration.
| Line | Cost |
|---|---|
| Discovery and scoping | $15,000 – $25,000 |
| Data remediation | $35,000 – $80,000 |
| Data 360 build | $25,000 – $70,000 |
| Agent build (3 agents) | $45,000 – $90,000 |
| Testing and governance | $20,000 – $40,000 |
| Enablement | $10,000 – $20,000 |
| First-year services total | $150,000 – $325,000 |
Plus roughly 0.5 to 1.0 FTE of ongoing admin capacity.
Scenario C: Enterprise multi-agent rollout
Profile: 1,000+ users, multiple business units, customer-facing and internal agents, multi-system identity resolution, formal compliance requirements.
| Line | Cost |
|---|---|
| Discovery and scoping | $40,000 – $80,000 |
| Data remediation | $100,000 – $250,000 |
| Data 360 build | $120,000 – $250,000 |
| Agent build (5+ agents) | $150,000 – $300,000 |
| Testing, governance, compliance | $60,000 – $120,000 |
| Enablement and change management | $30,000 – $60,000 |
| First-year services total | $500,000 – $1,060,000 |
Plus 1.5 to 2.5 FTE of ongoing capacity, often a dedicated AI operations function.
Across these scenarios, first-year services land at roughly two to five times first-year platform spend. If your business case assumed services were a rounding error on the license, the case needs rebuilding before you sign anything.
The Line Everyone Forgets: Ongoing Admin Load
An agent in production is an operated system, not a delivered artefact. The recurring work:
- Reviewing and approving agent-initiated changes
- Monitoring consumption against forecast
- Updating topics, instructions, and actions as the business changes
- Regression testing after prompt or Salesforce release changes
- Maintaining grounding content freshness
- Handling escalations when the agent gets something wrong
- Keeping the data clean enough that grounding stays reliable
Rough sizing: 0.25 FTE for a single internal agent, 0.5 to 1.0 FTE for a mid-market multi-agent deployment, 1.5 to 2.5 FTE at enterprise scale. At a loaded admin cost of $95,000 to $140,000, that is $25,000 to $350,000 a year of real operating cost.
Two ways teams get this wrong. The first is omitting it, which means it silently consumes an existing admin's capacity and something else stops getting done. The second is assuming the agent offsets it, which confuses two different pools of work. The agent adds throughput on repetitive execution. The operating load is judgment, review, and change management, and that scales with agent count rather than shrinking.
If you want a defensible number for your own org, the admin capacity calculator is a reasonable starting point, and Salesforce managed services is the alternative to hiring for it.
What Actually Moves the Number
In rough order of impact:
Org condition. The difference between a governed org and a post-acquisition sprawl is six figures on remediation alone. This is why assessment before quoting is non-negotiable.
Whether Data 360 is genuinely required. A correct "no" removes the second-largest line entirely.
Use case count in year one. Each additional agent adds cost, but sequencing them after the first is far cheaper than parallel delivery. Resist breadth in year one.
Integration surface. Every external system the agent touches adds build and testing effort, and external systems are where timelines slip.
Compliance requirements. Regulated environments add audit, documentation, and validation work that can be 20 to 30% on top.
Internal capacity. Teams that can do their own remediation with tooling and oversight pay significantly less than teams outsourcing it wholesale.
Controlling It
Buy the assessment separately. A paid, standalone assessment from a partner who is not guaranteed the build gives you an unbiased scope and a document you can competitively tender. It is the cheapest leverage available.
Ship one agent, properly. One agent with tested governance beats four half-built ones, and it makes agents two through four cheaper.
Insist on grounding-path justification. As above, this single question routinely removes the largest avoidable line.
Separate remediation from build commercially. Remediation scope is genuinely unknown before assessment. Fixed-pricing it means paying a risk premium. Time-and-materials with a cap and weekly reporting is usually fairer to both sides.
Budget handover explicitly. Make documentation and admin training a named deliverable with acceptance criteria, not a final-week activity that gets compressed.
Forecast and cap consumption from day one. Cheap to instrument up front, expensive to discover on an invoice.
The full set of questions to put to a vendor is in how to choose an Agentforce consulting partner.
How We Price It
Clientell is an AI-led Salesforce services team, so we sit on the services side of this equation and it is fair to be explicit about how that shapes our numbers.
We quote assessment separately and deliver it as a written findings document, including the grounding-path recommendation and a remediation estimate grounded in measured figures from your org rather than a template. You can take that document to any partner. Several clients have.
On remediation, our agent capability changes the cost structure honestly and in a specific way: the high-volume repetitive passes, duplicate identification, ownership mismatch detection, field population analysis, dead field discovery, are handled at a throughput a human team cannot match on the same budget, with your admin approving batches before anything is written. That compresses the largest and most variable line in the project. It does not replace the admin judgment about what should be merged or which fields matter, and we do not price it as though it does.
Build, testing, and governance are quoted per agent with rollback and audit trail as standard scope rather than options. Handover is a named deliverable.
If you want the assessment first, start with the Agentforce readiness audit or the Salesforce data quality audit if you already know data is the weak point. To scope a specific project, book a demo.
Frequently Asked Questions
How much does Agentforce cost to implement? First-year professional services typically run $40,000 to $80,000 for a single-use-case pilot, $130,000 to $280,000 for mid-market production, and $400,000 to $900,000 for enterprise multi-agent rollout, excluding licenses and consumption.
Is Agentforce implementation cost separate from license cost? Yes, entirely. Licenses and consumption are covered in the pricing guide. Services are typically two to five times first-year platform spend.
How long does an Agentforce implementation take? Sixteen to thirty weeks for a first production agent in a mature org. Data remediation, not the agent build, is usually the longest phase.
Do I need to pay for Data 360 as part of implementation? Only for unstructured grounding or cross-system identity resolution. Many internal use cases do not require it, which removes a large cost line.
What are Salesforce partner day rates in 2026? Roughly $400 to $800 offshore, $700 to $1,300 nearshore, $1,200 to $2,200 onshore boutique, and $2,000 to $3,500 for large systems integrators or Salesforce Professional Services.
Can we implement Agentforce without a partner? For a narrow internal agent in a clean, well-understood org, yes. The parts that usually justify outside help are the permission model review, the rollback and audit design, and remediation volume.
What is the ongoing cost after go-live? Plan 0.25 to 2.5 FTE of admin capacity depending on scale, plus consumption. This is a permanent operating cost and belongs in the business case from the start.