Agentforce Sales Coach: Editions and the Four Dependencies

Search for a Salesforce sales coach and you will not find a person. You will find a product: Agentforce Sales Coach, an autonomous agent that role-plays a customer, listens to a rep practise, and grades the attempt against the CRM record they were practising for.

It is a genuinely interesting piece of software, and it is also four dependencies deep before it does anything. Salesforce publishes both facts on its own pages, and the second one is the part a buyer needs first.

Everything below comes from Salesforce's own pages: the Sales Coach help documentation, the Agentforce pricing page and the Sales Cloud pricing page.

What is Agentforce Sales Coach?

Agentforce Sales Coach is Salesforce's coaching agent, the product that Salesforce sales coaching now means. It sits as a Lightning Web Component on any Salesforce object record, critiques a rep's presentation at the Qualification and Needs Analysis stages, and plays the customer at the Negotiation/Review and Proposal/Pricing Quote stages, grading each attempt against that record's CRM data. It needs a qualifying edition plus Foundations or an Agentforce 1 Edition.

What the agent actually does

Salesforce describes Sales Coach as an autonomous agent built on the Agentforce platform that uses generative AI and CRM data to provide personalised feedback, reachable on any Salesforce object record through the Agentforce Sales Coach Lightning Web Component. That last detail is the useful one: it is a component on a record page rather than a separate application, so a rep practises against the specific opportunity in front of them.

The coaching splits by opportunity stage, and the two halves behave differently.

The mechanism underneath both is the same, and it is what separates this from a generic chatbot doing sales roleplay. While analysing a pitch, presentation or role-play session, the agent cross-references what the user said against the CRM data for the record they were practising for. So a rep who claims a discount that does not match the quote, or who describes a use case the account record contradicts, is arguing with the system of record rather than with a model's general knowledge. Feedback can be augmented further using the Agentforce Data Library to reference external, company-specific data, which is retrieval-augmented generation applied to your own sales collateral.

Salesforce also notes that the agent can coach every customer-facing role in the organisation, not only sellers, which is a reasonable claim given the component sits on any object record.

The editions requirement, which is the first gate

Salesforce publishes the requirement plainly: Agentforce Sales Coach is available in Lightning Experience in Enterprise, Performance, Unlimited and Developer Editions, with Foundations or Agentforce 1 Editions.

The Sales Cloud pricing page, meanwhile, now sells Core, Advanced and Max rather than editions by those names, and says AI can be added to Core and above. Neither page maps the old edition names to the new tiers, so confirm with Salesforce which of your tiers the help article's list covers.

Read the requirement as two conditions rather than one. The org has to be on a qualifying Sales Cloud edition, and it separately has to have the Agentforce layer, whether through Foundations or through an Agentforce 1 Edition. A team on Professional Edition is not one setting away from this feature. They are an edition upgrade and an AI layer away from it, and the upgrade is the larger number.

The four things that have to be on underneath

Salesforce lists four required features for Agentforce Sales Coach to operate correctly, access CRM data and generate feedback.

None of those is exotic for an org already committed to the Salesforce AI stack. For an org that is not, the list is the honest scope of the project. Sales Coach is a feature of a platform build rather than a tool you switch on for a quarter to see whether coaching helps.

What it costs, and why no page will tell you exactly

Neither the Agentforce pricing page nor the Sales Cloud pricing page lists a Sales-Coach-specific price. What they publish is the licensing surface the feature sits inside, and the Agentforce pricing page now shows two different ways that surface can be paid for.

Per seat, unmetered. The page lists Agentforce add-ons for Sales, Service, and Field Service and Operations at $125 per user per month, described as "Unmetered Agentforce usage for employees", and an Industries add-on at $150. Agentforce 1 Editions start from $550 per user per month with the add-on included and 2.5M Flex Credits per org per year. The Sales Cloud pricing page's Max tier, also $550, lists 2.75M Flex Credits per org per year instead, so ask which figure applies to your contract.

Per action, metered. Salesforce Foundations is listed at $0 as a way to get started with Agentforce for free, and usage on that path is paid in Flex Credits. The page prices those at $500 per 100k credits, which is half a cent a credit, and says Agentforce actions are 20 Flex Credits and Agentforce Voice actions 30. At those rates one Agentforce action costs ten cents. The page's own worked example, a self-service agent answering 20 order requests a day at 2 actions each, comes to 40 Flex Credits, or US$0.20, per request and US$120 a month.

Which path an org is on changes what coaching costs. With a per-user add-on, repeated role-play by a licensed rep adds no metered cost. On the Flex Credit path, every coaching action draws from the pool, and a session is not one action. As illustrative arithmetic for ten reps, the add-on costs $1,250 a month, and at ten cents an action that sum buys 12,500 metered actions a month, or 1,250 per rep. A team that expects to coach below that line may be cheaper on credits; one above it is cheaper on seats.

The buying models apply to the meter, not the seats. The page lists Pre-Purchase, Pre-Commit and PayGo; Pre-Commit is marked as widely available later this year, and PayGo carries no upfront commitment and bills monthly in arrears. Its FAQ says Flex Credits can be bought on PayGo and Pre-Commit, while Agentforce add-ons and Agentforce 1 Editions cannot. There is no overage penalty: usage past the entitlement bills at the contracted rate, monthly in arrears, and Salesforce points at alerts in Digital Wallet as the way to prevent overages.

	Per seat	Per action
What you buy	Agentforce add-on for Sales, or an Agentforce 1 Edition	Foundations at $0, plus Flex Credits
Price on the page	$125 a user a month; Agentforce 1 from $550	$500 per 100k credits; 20 credits an action
Coaching usage	Unmetered for employees	Every action draws from the pool
Buying models	Not on PayGo or Pre-Commit	PayGo or Pre-Commit
Ten reps, illustrative	$1,250 a month	$1,250 buys 12,500 actions

What a sensible pilot looks like

The configuration surface is Agentforce Studio, which is where the subagents and their actions are set up, so the ready-to-use scenarios Salesforce ships are a starting point rather than the whole product. That cuts both ways. A team willing to edit the scenarios can point the agent at the objections their own market actually raises. A team that never opens Studio gets generic coaching wearing their CRM data, which is better than nothing and considerably less than the licence cost implies.

The thing worth deciding before the pilot is what would count as it working, because coaching resists measurement in a specific way. Comparing win rates before and after is almost always contaminated: the reps who use an optional coaching tool most are rarely a random sample of the team, so any lift you measure includes the fact that keen reps volunteered. A cleaner test is narrower. Take one recurring objection, coach only that, and count how often it still ends conversations three months later, with the population stated rather than filtered to the people who engaged.

1

Pick one recurring objection

One that ends conversations today, in your own market.

2

Edit the scenario in Agentforce Studio

Point the agent at that objection instead of the generic ready-to-use scenario.

3

Coach only that

Keep the test narrow enough to read.

4

Count three months later

How often it still ends conversations, with the population stated rather than filtered to the keen reps.

Where coaching helps, and the part it cannot reach

Coaching software has a narrow and real job. A rep who fumbles the pricing conversation on the same objection every time is a fixable problem, and rehearsing it against a system that knows what is actually on the quote is better than rehearsing it against a colleague who does not. The stage mapping Salesforce chose reflects that: the role-play modes sit on Negotiation and Proposal, which are exactly the conversations where a bad ten minutes costs a deal that was already won on merit.

Teams already weighing conversation intelligence against CRM-native coaching can see how the categories divide in the Gong and Salesforce comparison.

What none of it reaches is whether the conversation happens at all. An agent that grades a negotiation cannot create the meeting where the negotiation takes place, and a coaching layer applied to a pipeline that is too thin makes the reps better at a job they are not getting enough repetitions of. If the constraint is conversations per rep per week rather than quality per conversation, the spend belongs upstream. Our view of what actually moves that number is in what an AI sales agent can and cannot do, and the human version of the same argument is in SDR training.

A separate limitation worth noting is that coaching cannot create pipeline, and the comparison of sales engagement and CRM explains why outreach tools and Salesforce play different roles in the funnel.

Two practices of ours are worth naming beside a product like this, because they sit at the same boundary. Meetings are qualified against criteria agreed in writing before launch rather than judged after the fact, which is the same instinct as coaching against the record instead of against opinion, and the standard is set out in qualified appointment. And we run one message per campaign, with no bumps and no thread replies, so the outbound layer feeding those conversations produces one clean attribution per contact rather than a thread to untangle.

If you are weighing the platform decision underneath all of this rather than the feature, the Salesforce alternatives roundup covers the field, HubSpot against Salesforce covers the comparison most teams actually run, and sales engagement platforms covers the adjacent category that gets confused with it.

The short version

Agentforce Sales Coach is an autonomous agent on a record page that does two things: it critiques a rep's presentation at the Qualification and Needs Analysis stages, and it plays the customer at the Negotiation and Proposal stages, grading both against the CRM data for the record in question and, optionally, against your own documents through the Agentforce Data Library.

It requires a qualifying edition in Lightning Experience with Foundations or Agentforce 1, and it requires Agentforce, Einstein Generative AI, Agentforce Studio and Data 360 to be in place. Neither pricing page lists a price for the feature itself. Salesforce's pricing pages publish the Agentforce add-on for Sales at $125 per user per month with unmetered employee usage, the Industries add-on at $150, Agentforce 1 Editions from $550 per user per month with 2.5M Flex Credits per org per year, and Flex Credits on their own at $500 per 100k, which can be bought on PayGo or Pre-Commit.

Buy it if your reps are having enough conversations and losing them late. If the problem is the number of conversations, see what a campaign would look like for your market before you buy a coach for calls that are not being booked.

Pricing and features are taken from the vendors' own pages. Verify current terms with the vendor before relying on them.

Sources: Agentforce Sales Coach Overview, Salesforce Agentforce pricing