Salesforce publishes a price for a credit, a conversation, a user license, and a resolved case. Those numbers price the meter.
Your project is everything that happens before the meter starts running: defining the process, cleaning the records the agent will read, building and wiring actions, connecting the systems it calls, testing what happens when things go wrong, and deciding who owns it after launch.
So 2 companies can buy identical Salesforce SKUs and spend wildly different amounts getting an agent live. What separates them is the org, the process, and the design.
An Agentforce budget in 2026 has to account for all of it, and the sections below build that estimate one layer at a time.
TL;DR
Agentforce implementation cost splits into 3 layers that get budgeted, billed, and negotiated separately. Mixing them together is why published cost figures vary so wildly from one source to the next.
- Salesforce product charges. Flex Credits at $500 per 100,000, Conversations at $2, user licenses and add-ons, and resolution-based billing for the Help Agent. Published, predictable per unit, and easy to look up.
- Implementation services. Discovery, readiness remediation, architecture, agent build, integrations, Knowledge preparation, security, testing, deployment, and enablement. This is where budgets separate.
- Ongoing consumption and Data 360. Action volume, retrieval activity, Data 360 processing, monitoring, and change work after launch. Design decisions made during the build set this number for years.
- Year-one total and cost control. Add the 3 layers, then model action count before you sign anything, because action design drives the recurring bill more than user count does.
What Is Included in Agentforce Implementation Cost?
Ask 3 vendors what “implementation cost” covers and you’ll get 3 different scopes. Define the layers first, and the numbers start meaning something.
Salesforce product cost
Fees paid to Salesforce for access and usage. That covers Flex Credits or Conversations, any user licenses or add-ons the agent type requires, and outcome-based charges where the product supports them.
Implementation services cost
Professional work to get a defined use case into production. Discovery, configuration, custom development, integration, Knowledge preparation, security design, testing, and deployment all sit here.
Operating cost
What it costs to keep the agent running. Consumption, Data 360 activity, monitoring, support, and the change work every new release or new use case creates.
| Cost Layer | What It Pays For | Billing Pattern |
|---|---|---|
| Salesforce | Product access and consumption | Subscription and usage |
| Implementation | Project work to reach production | Project or services agreement |
| Operations | Running and maintaining the agent | Recurring |
Keep these separate on the budget line. Our full Agentforce cost breakdown including Data 360 goes deeper on the consumption side if that’s the layer you’re modeling first.
Why Agentforce Implementation Costs Vary So Much
2 projects can target the same headline outcome and require completely different effort. The gap opens because implementation work tracks the system being built.
Picture it as a chain: business scope, then technical design, then implementation effort, then Agentforce consumption, then ongoing cost. Each link sets up the next.
Business scope decides how many processes get analyzed, how many stakeholder groups sign off, and how many edge cases the agent has to handle. A single bounded workflow needs a fraction of the discovery a multi-department program does.
Technical design turns that scope into components. Every action, integration, and data source becomes something to build, secure, and test.
Implementation effort follows from the component count and the condition of what’s already in the org. A clean Flow you can reuse costs less than a tangle of overlapping automation someone has to unpick first.
Design decisions then carry forward into consumption, because Agentforce bills per action. Budgets routinely stop at the build and miss this link entirely.
A workflow built to complete in 3 actions costs less every time it runs than the same outcome built in 6, and that difference compounds across every request for the life of the agent.
Here’s how it plays out. 2 companies both want an agent answering order-status questions, and the first reads clean order records already sitting in Salesforce.
The second holds order data in an ERP, so the same answer needs an integration, credential management, failure handling, and tests covering what the agent says when the call fails. Same outcome, same headline use case, and an integration project instead of a configuration exercise.
Salesforce Agentforce Pricing in 2026
Here’s what Salesforce publishes today. Check the live pricing page before you contract, because these terms move between releases.
| Pricing Route | Current List Price | Billing Unit | Typical Use Context |
|---|---|---|---|
| Salesforce Foundations | $0 | Included entitlement | Early experimentation with included Flex Credits, restrictions apply |
| Flex Credits | $500 per 100,000 credits | Credit consumed | The default consumption model for most agent work |
| Conversations | $2 per conversation | Conversation | Customer-facing agents billed per conversation window |
| Agentforce User License | $5 per user per month | User, plus Flex Credits | Metered employee usage where a full add-on isn't needed |
| Agentforce add-ons | $125 per user per month | User | Unmetered Agentforce usage for employees |
| Agentforce Industries add-ons | $150 per user per month | User | Industry cloud equivalents |
| Agentforce 1 Editions | From $550 per user per month | User, bundled entitlement | Includes 1M Flex Credits and 2.5M Data 360 Credits per org per year |
| Help Agent Resolutions | $2 per resolution | Successful resolution | Self-service deflection billed on outcome |
3 details matter more than the numbers themselves, and each one causes budget errors when it’s missed.
First, the Agentforce 1 entitlements are 2 separate pools. 1 million Flex Credits and 2.5 million Data 360 Credits aren’t interchangeable, and a page quoting a single figure has merged them.
Second, Salesforce consolidated its US Sales and Service edition lineup in September 2026, so edition names on a 2025 quote may not match what’s sold today. Third, the Foundations credit figure is inconsistent on Salesforce’s own page: the feature list shows 450K Flex Credits shared with Data 360, while the FAQ on that page says 200K. Confirm which applies to your account before planning a proof of concept around it.
One caveat rides on the add-on rows. “Unmetered” describes the user-facing agent activity, and it depends on the user’s licenses, permissions and the specific functionality. Underlying services can still draw credits, so a Data 360 query behind an unmetered employee interaction is still billable.
Our Salesforce Agentforce pricing guide covers the product side in full, and Salesforce’s own pricing page is the authority on list rates. Salesforce also notes that pricing is subject to change and that contract-specific terms come from your account team, so treat every figure here as a starting point for a conversation. Everything in this table is money paid to Salesforce, with implementation work costed separately.
How Agentforce Flex Credits Turn Usage Into Monthly Cost
This is the calculation that turns a use case into a monthly number.
Understand the billing unit
Flex Credits cost $500 per 100,000, which works out to $0.005 per credit. A standard or custom action in production consumes 20 credits, so Salesforce prices one action at $0.10.
Sandbox is cheaper for actions and identical for prompts. Actions drop to 16 credits, voice actions run 30 in production and 24 in sandbox, and prompts cost the same in both (starter and basic at 2, standard at 4, advanced at 16).
Voice carries one trap worth naming. Alongside voice actions, Salesforce runs a separate Voice Minutes model on a supplemental rate card that bills per rounded-up minute instead of per action. Confirm which unit your contract uses before applying any text-action arithmetic to a voice deployment, because the 2 models produce very different numbers on the same call volume.
One fact does more work than any other when you’re forecasting. Salesforce documents that “every standard action called by the agent is metered as one standard action, irrespective of the number of tokens processed.” Action count drives the bill, so a long exchange costs the same as a short one at equal actions.
Use the monthly usage formula
Monthly action cost = monthly business events × billable actions per event × 20 credits × $0.005
At current list pricing that collapses to something you can do in your head:
Monthly action cost = monthly events × actions per event × $0.10
Show why action design matters
Salesforce publishes its own worked examples, and they make the point better than an invented scenario.
| Salesforce Example | Events Per Month* | Actions Per Event* | Credits | Monthly Cost |
|---|---|---|---|---|
| Employee onboarding questions | 100 | 1 | 2,000 | $10 |
| Case management | 6,000 | 3 | 360,000 | $1,800 |
| Field service appointments | 600 | 6 | 72,000 | $360 |
Credits and monthly totals come straight from Salesforce. The starred columns restate its volumes (100 users at 3 cases a day across 20 days) as one monthly figure.
Compare the middle row with the bottom one. Field service runs 10% of the case volume at 6 actions against 3, so it costs a fifth as much where volume alone suggests a tenth.
If you want the mechanics behind why a given request consumes the actions it does, our guide to how Agentforce works walks through the reasoning loop, and Salesforce’s Flex Credits rate card lists every multiplier.
One boundary on this math: it covers Agentforce action consumption. Licensing, Data 360 activity, implementation work, and ongoing support are all separate lines.
Compare Agentforce Pricing Models Before Choosing One
Salesforce sells access to the same technology several ways, and the right one depends on your usage pattern rather than a universal cheapest option.
| Question | Flex Credits | Conversations | Outcome-Based | User or Add-On |
|---|---|---|---|---|
| Billing unit | Credit consumed per action | Conversation window | Successful resolution | Licensed user per month |
| Forecast input | Events × actions per event | Conversation starts | Resolution rate | Headcount |
| Best-fit audience | Mixed agent workloads | Customer-facing messaging | Self-service deflection | Employee-facing agents |
| Usage variability | Cost tracks volume directly | Cost tracks contact frequency | Cost tracks success only | Cost is flat regardless of use |
| Contract consideration | Unused credits expire with the term | Cannot coexist with Flex Credits | Unresolved sessions aren't billed | Per-seat commitment |
3 rules constrain the choice. Salesforce states that “Flex Credits and Conversations will not be supported in the same org,” making this an org-level decision rather than a per-agent one.
Unused Flex Credits “do not roll over into subsequent subscription terms,” so over-buying is a real loss while under-buying bills the overage at your contracted rate. Forecasting belongs before contract sizing for that reason.
Payment shape stays negotiable even when the unit rate holds firm. Salesforce currently publishes pay-as-you-go, where usage bills monthly in arrears with no upfront commitment, and pre-commit, where you commit to baseline usage without paying upfront and true up against the contract. Buying a fixed quantity upfront for the term is the third arrangement to ask your account team about, since the published page names the first 2 explicitly.
Outcome-based billing is worth understanding even if you don’t choose it. The Help Agent charges only when it resolves an issue end to end, and negative feedback or a human escalation means no charge. Its rate card row sits at 400 credits, which reconciles exactly to the published $2.
Customer-facing deployments carry different economics from employee ones, which is why Agentforce for Service work gets scoped differently from an internal rollout. Salesforce’s pricing documentation sets out the current unit rates.
What Agentforce Implementation Services Actually Cover
Published implementation ranges conflict because each page defines the work differently. One counts agent configuration. Another counts data cleanup, integration, training, and a year of support.
Price the workstreams your project actually needs. The 12 below are what a full Agentforce delivery contains, and an estimate is the sum of whichever ones apply to you.
| Workstream | Work Involved | What Changes the Effort |
|---|---|---|
| Discovery | Process and use-case definition | Number of processes and stakeholder groups |
| Readiness | Org, data, and automation review | Volume of technical debt found |
| Architecture | Data, action, and system design | Count of systems in scope |
| Agent build | Instructions, subagents, actions | Use-case complexity |
| Development | Flow, Apex, and custom actions | How much existing automation is reusable |
| Integration | External system connections | API condition and call volume |
| Knowledge | Content cleanup and preparation | Quality of the source material |
| Security | Permissions and access controls | Compliance requirements |
| Testing | Functional and failure scenarios | Channel and action count |
| Deployment | Environments and release work | Number of channels and orgs |
| Enablement | Procedures and training | Number of user groups |
| Support | Monitoring and change work | Coverage hours required |
A quote covering 4 of these 12 is simply a smaller quote. The remaining 8 get absorbed by your team or show up as change orders later.
Some should be absorbed internally, and that’s a legitimate way to control spend. Enablement and first-line support often sit better with the team who’ll live with the agent, and readiness work does too if you have an admin who knows where the problems are buried.
Architecture, security design, and failure testing rarely transfer well. Decide which column each workstream lands in, then hold every proposal to that split.
Scoping runs against the actual condition of a Salesforce org and the specific use case in front of it, which is why two estimates for “one service agent” can land in completely different places. The VALiNTRY360 Agentforce consultant guide sets out what that scoping conversation should produce on paper.
8 Factors That Change Agentforce Implementation Cost
- Use-case scope. A bounded workflow with one trigger and one success condition can be specified in days. A program spanning sales, service, and operations multiplies the analysis, approvals, and test matrix.
Cost effect: the largest swing factor, since it sets the size of every other workstream.
- Number of agents and subagents. Each additional job brings its own instructions, actions, escalation paths, and test cases. Effort climbs steeply with each distinct job, even where agents share design patterns.
Cost effect: design and testing effort grows with each distinct job.
- Salesforce org condition. Technical debt surfaces during implementation whether you budgeted for it or not. Duplicate automation, undocumented customizations, and stale records become remediation work before the agent can be trusted.
Cost effect: unpredictable until someone audits the org, which is why readiness reviews come first.
- Number and type of actions. Standard actions configure quickly. Custom Apex, composite actions enforcing a fixed sequence, and API-backed actions each carry development and test time.
Cost effect: raises build cost once, then raises the consumption bill on every run.
- Integrations. Every ERP, billing platform, commerce system, or legacy API adds authentication design, error handling, latency testing, and failure behavior.
Cost effect: often the difference between a 6-week pilot and a 6-month program.
- Data and Knowledge condition. If the records or articles the agent grounds on are wrong, duplicated, or unowned, cleanup becomes its own project ahead of the build.
Cost effect: can rival the agent build when content has been neglected for years.
- Security and compliance. Regulated processes bring permission design, approval gates, audit trails, sensitive-data review, and injection testing, plus the cycles to sign all of it off.
Cost effect: adds elapsed time as much as effort, because approvals queue.
- Channels and operating model. Web chat, service channels, Slack, voice, and headless deployments each need their own configuration and testing. Voice also bills on different multipliers.
Cost effect: each additional channel repeats the deployment and test workstreams.
Integration is the factor teams underestimate most consistently, since that connection layer decides whether everything built on top of it holds. Agentforce integration work is scoped as its own stream for exactly that reason.
Data 360, Knowledge, Integrations, and Custom Actions Can Change the Budget
Data 360 cost
Whether Data 360 costs you anything depends on your architecture. Agentforce Data Library needs it, and which credit pool pays depends on your contract: Salesforce states that if your contract included Data Service Credits, those are consumed for all Data 360 usage, and if you hold Flex Credits for Data 360, those get consumed instead.
The rate card is tiered, which matters for forecasting. Queries run 3 credits per million rows at the base tier and fall to 0.6 at the top, while unstructured processing runs 150 credits per megabyte at base and 30 at the top.
Sandbox rates are flat, with no tiering.
So unit cost drops as volume rises, and a pilot-sized estimate scaled linearly to production overstates the bill. Sizing this properly means working from the processing types a use case actually triggers, which is the approach behind Data 360 scoping, and Salesforce documents the Data Library billing rules directly.
Knowledge preparation
Article cleanup, duplicate removal, assigning owners, reviewing expiry dates, and preparing documents for indexing. Work here is proportional to how long the content went unmanaged.
Knowledge carries a risk CRM records don’t. The agent treats an article as a finished answer, so a contradictory pair produces a confidently wrong response. Content review is a correctness requirement here.
Scope it to the questions the first agent will field rather than the whole library. A support team with 800 articles usually finds that 60 of them cover the bulk of what the pilot needs to answer.
Integration cost
Authentication, API development, error handling, middleware configuration, and testing what the agent does when a connection fails. The last one gets skipped and then gets expensive.
Budget for latency and rate limits specifically. An API that answers comfortably for a nightly batch job behaves differently when an agent calls it mid-conversation, and finding out after go-live means rework.
Ownership matters as much as engineering here. If the system on the other end belongs to another team or another vendor, their release calendar becomes part of your timeline.
Custom action cost
Flow and Apex development to make an action reliable, plus the recurring credit consumption every time that action fires.
Build quality and run cost meet here. Consolidating 3 loose actions into 1 composite action cuts the sequencing risk and 40 credits off every execution.
At 50,000 executions a year that single design decision is worth $10,000 annually, which is usually more than the extra build time cost in the first place.
| Technical Dependency | Project-Cost Effect | Run-Cost Effect |
|---|---|---|
| Data 360 | Configuration and data preparation | Tiered Data 360 consumption |
| Knowledge | Content cleanup and indexing | Retrieval activity |
| Integration | Build and failure testing | API and platform usage |
| Custom action | Flow and Apex development | 20 credits per production action |
Read that table as a warning about sequencing. A dependency that looks cheap in the build column can still be the expensive one over 3 years, because the right-hand column bills every month while the left-hand column bills once.
Testing, Security, Deployment, and Adoption Also Belong in the Implementation Budget
These 4 sit at the end of the project plan, and they belong near the top of the estimate. Work through them before you approve a number.
Testing. Component tests, action-level tests, conversation tests, negative cases, integration failure simulation, business UAT, and a regression set you can rerun after every change. Negative cases carry most of the value and most of the effort, since an agent that only ever sees clean inputs during testing has barely been tested.
Security. Permission design against least privilege, sensitive-data review, approval controls on high-impact actions, and prompt-injection testing. Regulated processes add review cycles that queue behind other people’s calendars.
Deployment. Sandbox strategy, target environment preparation, confirming required Flows are active, credentials, version management, and channel setup. Add a documented rollback path and a named person who can deactivate a bad version.
Adoption and operating procedures. User guidance, escalation procedures, named change ownership, and a support process that exists before the first incident. Employee-facing agents need more of this, since staff have to learn when to trust the agent and when to take over.
One 2026 change works in your favor here. Salesforce states that “as of Summer ’26, testing in Agentforce Testing Center is unmetered and doesn’t consume Einstein Requests or Flex Credits,” though “Data 360 queries from Agentforce Testing Center are still metered.” Verify the current wording before you commit a testing figure, because this is exactly the kind of term that shifts between releases.
Sandbox actions also run at 16 credits against 20 in production, so a heavy test cycle costs less than the same volume in production, while prompt-heavy testing costs the same in both. Our Agentforce testing guide covers what a real test plan contains, and Salesforce’s Testing Center considerations carry the current metering rules.
Pilot Versus Enterprise Rollout, and In-House Versus Implementation Partner
Pilot versus enterprise implementation
| Cost Dimension | Controlled Pilot | Enterprise Rollout |
|---|---|---|
| Processes | 1 bounded workflow | Several across functions |
| Agents | 1 agent, few subagents | Multiple agents with shared design |
| Integrations | Minimal or none | Several enterprise systems |
| Data scope | Records for one use case | Shared architecture across domains |
| Security | Single permission model | Regional and regulatory variation |
| Test coverage | Core paths plus key failures | Full matrix across channels and agents |
| Governance | Named owner | Formal operating model and change board |
| Support | Business-hours monitoring | Defined coverage and incident process |
Enterprise programs cost more because each added function brings another process to document, another set of records to trust, another approval chain, and another column in the test matrix. Holding a first use case inside a deliberately narrow boundary is what an Agentforce Quickstart scope looks like in practice.
In-house versus implementation partner
Compare honestly across available Salesforce skills, architecture experience with agent design, custom development capacity, integration capability, test coverage your team can realistically produce, and who owns the agent after go-live.
An internal team that runs Flow and Apex confidently, with capacity to spare, can deliver a bounded pilot. Outside help earns its place when the architecture spans several systems, when compliance review is involved, or when your team has yet to take an agent through a production failure. We wrote about whether you still need a partner for Agentforce in more detail, including when the honest answer is no.
Build Cost Versus Run Cost: Calculate Year-One Agentforce TCO
Budgets break when build and run get added into one figure, then compared against another company’s number that split them differently.
This is the same money the three-layer table at the top of the article describes, cut a different way. Salesforce charges and operations both sit in run, while implementation sits in build.
Build cost
Everything in the workstream table that ends when the agent goes live, paid once per use case. The heaviest lines are usually remediation and integration, since both depend on conditions you inherit rather than choices you make. Scoping that work is the same discipline behind any Salesforce implementation services engagement.
Run cost
Flex Credits or the applicable usage model, user licenses and add-ons, Data 360 consumption, monitoring, support, change requests, and regression testing after each release. Paid every month, forever.
Year-one Agentforce TCO = Salesforce charges + implementation services + remediation + operating cost
Year-two run rate = recurring Salesforce charges + ongoing support + expected change work
4 steps get you to a defensible number:
- Estimate monthly events and actions per event, then price consumption using the $0.10 per action math from earlier. Build a low and high version, since actions per event is the least certain input you have. 2. Add the Salesforce access charges your agent type requires, and check what your current contract already covers before buying anything new. 3. Price the implementation workstreams in scope, and mark the ones you’ll absorb internally. Internal effort is still cost, it just lands on a different line. 4. Add Data 360, support, and the change work you expect in the first 12 months.
Year two looks different. Build cost falls away, consumption rises with adoption, and change work replaces implementation work. Teams who model only year one get surprised by a second year that’s cheaper in total but far more variable.
Not every category applies to every organization. A bounded employee agent on existing licenses with clean data skips several lines entirely. Salesforce’s usage and billing documentation covers how each usage type is measured.
Use This Agentforce Cost Estimator Before Requesting a Quote
Fill this in before you talk to anyone. Vendors price work far more accurately when these inputs arrive with the request, and you’ll spot scope gaps in competing proposals immediately.
2 inputs carry more weight than the rest. Actions per transaction sets your recurring bill, and external systems sets your build risk. If you’re unsure of either, say so in the brief, because a vendor pricing against a confident wrong number quotes confidently and wrongly.
Cost-estimation worksheet
| Input | Your Planning Value |
|---|---|
| Business process count | |
| Agent and subagent count | |
| Monthly transaction volume | |
| Actions per transaction | |
| Existing Flows reused | |
| New custom actions | |
| External systems | |
| Knowledge sources | |
| Data 360 requirements | |
| Channels | |
| Security requirements | |
| Test scope | |
| Support requirement |
Quote-comparison checklist
Ask every proposal to state clearly whether it includes each of these. Where two quotes differ by a large multiple, the answer is usually sitting in this list rather than in the day rate.
Discovery, readiness assessment, data remediation, Agentforce configuration, Flow and Apex development, integrations, Knowledge preparation, Data 360 configuration, security design, testing, deployment, training, post-launch support, Salesforce fees, and estimated ongoing consumption.
The last item trips people up. A proposal can be complete on services and still leave you holding an unmodeled consumption bill.
When two quotes differ sharply, run both against this list before assuming one vendor is expensive. Usually one priced remediation and failure testing while the other quietly assumed clean data and well-behaved APIs.
Salesforce’s own Agentforce pricing calculator states plainly that “Salesforce licenses, Data 360 credit cost, and implementation costs not included.” Salesforce’s calculator prices consumption. This worksheet prices the project around it.
This checklist covers what a proposal must state, while the workstream table earlier covers what the work itself contains. Read together they answer two different questions: what you’re buying, and whether the quote in front of you actually prices it. VALiNTRY360 sets out the structure behind that kind of scoping under Salesforce consulting services.
How to Control Agentforce Cost and Decide Whether the Investment Makes Sense
Cost-control actions
Start with one measurable process. A bounded use case produces an estimate you can defend and a baseline you can measure against.
Model action count before launch. Consumption follows actions per event. Get that right and the forecast holds.
Reuse tested automation where it fits. Flow and Apex that already works is cheaper to call than new development.
Limit remediation to what the use case needs. Clean the records the first agent reads, not the whole CRM.
Keep fixed rules in deterministic logic. Sequences that must run the same way every time belong in a Flow or Apex action rather than natural-language instructions.
Replace estimates with pilot data. Once real consumption exists, the forecast stops being a guess.
Watch consumption in Digital Wallet. It gives near real-time usage by agent and action, plus threshold alerts. It reports rather than caps, so hitting 100% of an entitlement triggers no automatic stop and budget control needs a named owner reviewing it.
Recalculate after major changes. New actions, channels, or data sources change the run rate.
Measure the business case
Estimated annual benefit − annual Agentforce TCO = estimated net benefit
What counts as benefit depends on the use case. Labor time avoided, volume handled without a person, self-service outcomes, employee capacity returned, cycle time reduced, and service cost per completed task are all candidates. Pick the one your pilot actually moves and measure it against a baseline captured before launch.
Skip universal ROI percentages. Salesforce says its own estimates “may vary based on architectural design, data requirements, and specific use case,” and that caution applies to any number you’re handed.
The honest version of this calculation carries a range rather than a point. Model the benefit at a conservative adoption rate and again at the rate your business case assumes, then look at how far apart the two net figures sit. A use case that only works at the optimistic end is worth knowing about before the build starts rather than after.
Scaling past the first agent changes the arithmetic again, which the enterprise Agentforce implementation guide covers in more depth. Once you’re live, Salesforce’s Digital Wallet is where the real consumption figures show up, and those replace every estimate on this page.
Salesforce Implementation Case Studies Worth Reading Before You Budget
Cost drivers are easier to judge against real projects than against a worksheet. If you want to see what the workstreams above look like when they’re actually delivered, these 5 VALiNTRY360 implementations each demonstrate a different driver this article covers.
- Real California Milk: 500 Annual Hours Cut to 4
Real California Milk, an agricultural marketing board, moved off SugarCRM onto Salesforce Sales Cloud. The project covered Account and Contact migration, an ETL tool with custom automations handling quarterly imports and reconciliation, plus MailChimp and Zoom integrations and custom management dashboards. Annual labor hours spent manually updating data fell from 500 to 4.
Key takeaway: Data migration and reconciliation logic is its own workstream with its own budget line, and the payback shows up as recovered staff time rather than as a lower platform bill.
Read the full case study: Real California Milk SugarCRM to Salesforce Migration
- AthenaPsych: 69% Fewer Coding Errors After an EMR Replacement
AthenaPsych, a mental health provider operating across New York State, replaced its Electronic Medical Record system outright with a Salesforce build on Health Cloud, Scheduler, Shield and Experience Cloud, integrating Change Healthcare and Vonage. Within 6 months of phase 1, 80% of patients were scheduled on the first call, 68% completed intake within 7 days, CPT coding errors fell 69%, and clinical compliance errors fell 75%.
Key takeaway: Replacing a system rather than integrating alongside it raises build cost and compliance review, and regulated work adds approval cycles to the timeline as well as effort.
Read the full case study: AthenaPsych Health Cloud EMR Replacement
- BIC Graphic: 27% More Pipeline and 7 Minutes Off Handle Time
BIC Graphic, a promotional products manufacturer operating at scale, had no CRM and fragmented customer data across sales, marketing and service. The implementation covered Sales Cloud and Service Cloud together with an integration into the existing ERP for order and fulfillment data. Average handle time fell by 7 minutes and new pipeline opportunities rose by more than 27%.
Key takeaway: Multi-cloud scope plus an ERP integration is the profile that separates an enterprise rollout from a pilot, and it’s where the integration and test workstreams grow fastest.
Read the full case study: BIC Graphic Sales and Service Cloud Implementation
- All American Solar: 16% Quarterly Growth From a Quickstart
All American Solar, a solar distributor and installer, needed Salesforce running without a dedicated administrator. The Quick Start covered lead and opportunity process mapping, record types and roles, Lightning for Gmail synchronization, dashboards for installation and KPI tracking, and 4 hours of on-site training. The company reported 16% quarterly growth year over year afterwards.
Key takeaway: A deliberately bounded scope with training built in is the cheapest path to a working system, which is the same logic behind scoping a first agent to one measurable process.
Read the full case study: All American Solar Sales Cloud Implementation
- AdventHealth University: 25 Years of Records Consolidated
AdventHealth University needed 25 years of alumni records, scattered across departments, pulled into one system. The build covered department-level access controls, a full ETL migration, opportunity and activity tracking, Outlook integration for automatic activity logging, plus testing and user training. Staff gained a complete view of each alumnus after graduation.
Key takeaway: Historic data consolidation and permission design are frequently the two workstreams that carry the most unbudgeted effort, especially when records have accumulated across separate teams.
Read the full case study: AdventHealth University Salesforce Implementation
Every one of these ran on a defined scope with a measurable outcome attached, which is exactly what the estimation worksheet above is built to produce before a project starts.
FAQs About Agentforce Implementation Cost
How much does Agentforce implementation cost in 2026?
There’s no universal figure. Salesforce product charges are published per unit, while implementation cost depends on use-case scope, org condition, integrations, data quality, and test coverage.
How much does Salesforce Agentforce cost per month?
It depends on the pricing route. Flex Credits cost $500 per 100,000, Conversations are $2 each, add-ons run $125 per user per month, and Agentforce 1 Editions start at $550 per user per month.
How much does one Agentforce action cost?
A standard or custom production action consumes 20 Flex Credits, which Salesforce prices at $0.10 at current list rates.
What are Agentforce Flex Credits?
Salesforce’s consumption currency for Agentforce, sold at $500 per 100,000 credits, or $0.005 each. Different usage types consume different amounts.
How many Flex Credits does an Agentforce action use?
20 credits for a standard or custom action in production, 16 in sandbox. Voice actions consume 30 in production and 24 in sandbox.
Does Agentforce still cost $2 per conversation?
Yes. Salesforce continues to publish Conversations at $2 for customer-facing agents alongside the Flex Credits model.
Can Flex Credits and Conversations be used in the same Salesforce org?
No. Salesforce states that Flex Credits and Conversations won’t be supported in the same org, making this an org-level decision.
How much does the Agentforce User License cost?
$5 per user per month, and it requires Flex Credits separately for metered usage.
What is included with Salesforce Foundations for Agentforce?
Foundations is a $0 add-on with included Flex Credits. Salesforce’s page currently lists 450K shared with Data 360 while its own FAQ says 200K, so confirm the figure for your account.
Does Data 360 increase Agentforce implementation cost?
It can, depending on architecture. Agentforce Data Library requires it, and Salesforce bills that usage from Data Service Credits or Flex Credits based on your contract.
Does Agentforce Testing Center add to implementation cost?
As of Summer ’26, Testing Center tests are unmetered and don’t consume Einstein Requests or Flex Credits. Data 360 queries triggered from those tests are still metered.
Which factors increase Agentforce implementation cost the most?
Use-case scope, org condition, and integration depth move the number furthest. Data and Knowledge remediation can match the agent build when content has been neglected for years.
How can I estimate Agentforce usage before implementation?
Multiply expected monthly events by billable actions per event, then by $0.10 at current list pricing. Refine it with real consumption data once a pilot runs.
Is it better to implement Agentforce in-house or use a Salesforce partner?
An internal team with Flow and Apex capacity can deliver a bounded pilot. Outside help earns its place with multi-system architecture, compliance review, or no internal experience of an agent failing in production.
What should an Agentforce implementation quote include?
Discovery, readiness, data work, configuration, Flow and Apex development, integrations, Knowledge, Data 360 setup, security, testing, deployment, training, post-launch support, Salesforce fees, and estimated ongoing consumption.