Costing for AI Agents Pilot and Agentforce Pricing Simplified

The foundation of any AI Agent pilot is, identification of suitable use cases for agents to perform their actions on. A good practice is to carefully consider the complexity of the use cases – not too complex as to create the risk of implementation cycle time and costs running away beyond control and not too easy as to run the risk of rendering the agents’ impact insipid.

AI Agent and Its Costs at a Glance

An AI Agent pilot has following costs:

1. Assessment Cost (if you choose to get it done by experts), wherein an AI Expert assesses your Business and IT Landscape and identifies the Use Cases as well as AI Agent Platform for you.

2. Platform licenses and AI Agents’ running cost

3. AI Agents pilot implementation cost

Here we are not going to account for platform license cost e.g. Salesforce Sales/Service/Marketing Cloud costs which is a sunk cost in most of the cases. However, if you are not using any such platform then you should account for this cost too.

In this paper we are assuming that one has already chosen, as an AI Agent platform, the Agentforce, from Salesforce.com stable.

Just to reiterate, an AI Agent interacts with the end users in a conversational style and executes the tasks as instructed by the end user in the conversation.

Agentforce platform provides predefined or customized actions tied to a topic to keep the definition and scope of an agent clear and to minimize the risk of Agent overstepping its mandate.

An Agent uses LLM, reasoning, memory, and platform tools (in this case Salesforce platform) to execute the instructions in the conversation. During execution an AI Agent could make several calls to LLM and Reasoning Engine, thereby consuming tokens which constitute a large part of the running cost. Hence, it is a good idea to have an estimate of this cost not just for the pilot but also overall because it is easy for this cost to remain obscure and balloon up quickly once the Agent is deployed and end users start using it in their day-to-day work.

If you have the estimates ready, you can choose the right pricing model and negotiate a better deal with the AI Agent platform vendor.

AI Agent platforms like Agentforce and others set their pricing keeping in mind potential calls to LLMs apart from many other things.

Broadly speaking there are three ways AI Agent Platforms set the pricing –

1.      Action Based – based on number of actions an Agent is performing

2.      Conversation based – based on conversations users are having with the AI Agent

3.      Per User Licensing – based on per user per month with some upper limit on free

consumption credits available and then additional cost per credit consumed.

Credits could be purchased in bundles to be cost effective based on your estimate of potential usage

For ease of understanding let us pick up two simple use cases – 1. from customer service the use case selected for pilot may be – provide and log the summary of the just finished conversation with the customer.  2. The other example could be from sales – send automated nurturing emails to potential customers.

In consumption-based model, pricing is done per action rather than per token, however, if for an action, tokens cross the limit of 10,000 then that would be counted as additional action(s), in other words if an action takes 20001 tokens, then it would be counted as 3 actions. Each action costs, on an average, 20 flex credits which translates to 0.10 cents approximately.

	Description
AI Agent Name	Customer Service Agent
Role	Summarize the customer interaction and log the summary in a field on Case object
Action	Provide and log the summary of a customer interaction
Number of Users (X)		5
Customer Interactions/User/month (Y)		5000
Tokens consumed/customer interaction (W)	Assuming 375 words per conversation and 0.75 words per token consumed	500
Token Consumption/Month C = (X*Y*W)		12,500,000
# of actions A	C/10000	1250
Price per action P	0.1 USD
Cost/Agent/month A*P		125 USD
	Description
AI Agent Name	Sales Development Representative
Role	Nurture Assigned Leads
Action	Send automated nurturing emails to leads
Number of Users (X)		5
Leads/User/month (Y)		200
# of emails / lead/month (Z)		100
Tokens consumed/email	Assuming 150 words per email and 0.75 words per token consumed	200
Token Consumption/Month C = (X*Y*Z*W)		20,000,000
# of actions A	C/10000	2000
Price per action P	0.1 USD
Cost/Agent/month A*P		200 USD

The above table provides two examples of going about estimating the running cost of AI Agents. This helps in selecting the right pricing model for your pilot.  Scope of a pilot is always limited to few use cases with limited number of users which warrants (often) going for a consumption-based pricing model instead of a fixed pricing model (per user/month with default credits).

Once pilot is deemed successful and you have plugged in your numbers for an organization wide implementation you might find that you require a different pricing model. You can negotiate a significant volume discount if you have all the numbers in hand.

Here is the Agentforce pricing info from Salesforce.com site:

Salesforce Agentforce Pricing (as per Salesforce.com)

Pricing Models Offered

Salesforce gives multiple ways to pay for Agentforce:

• Flex Credits (consumption-based)

• Conversations (per conversation)

• Per-user licensing / add-ons / editions

Key Price Points & Options

Model / Option	Price	Notes / What’s Included
Flex Credits	US $500 per 100,000 credits	Works based on “actions” that agents perform; you draw from the credit pool
Per Conversation	US $2 per conversation	For customer-facing use cases
Agentforce Add-On (employee use)	US $125 per user / month	Unmetered usage for employees (Sales, Service, Field Service)
Agentforce Industries Add-On	US $150 per user / month	For Industry Clouds (with industry-specific AI)
Agentforce 1 Editions	From US $550 per user / month	Comes with 1 million Flex Credits / year, etc.
Agentforce User License	US $5 per user / month	Requires Flex Credits (metered)

• In August 2025, Salesforce introduced pay-as-you-go and pre-commit models to its pricing structure.

• Under the Flex Credits model, each “action” costs 20 Flex Credits, which is roughly US $0.10 per action.

• The conversation model and Flex Credits cannot be used simultaneously in the same org.

Agentforce Price for German market and Comparison with other vendors:

Here are estimates showing prices converted into EUR for Germany (using current USD → EUR rate) + adding German VAT (19%), plus example monthly costs for 5-agent and 50-agent scenarios. These are rough estimates meant to give directional comparison; actual contracted prices may vary. Also this doesn’t take into account the token/action volume beyond the default volume allowed with the listed price.

We strongly recommend not to choose an AI Agent platform solely based on list price.

Assumptions

Parameter	Value
USD → EUR exchange rate	USD 1 ≈ EUR 0.88 (Forbes)
German VAT rate	19% for most goods/services. (All About Berlin)

Converted Prices + VAT

Here are approximate converted prices for some of the offerings we compared, inclusive of VAT:

Vendor / Plan	Original USD Price	Converted to EUR (≈ × 0.88)	Plus 19% VAT → Estimated EUR Price incl. VAT
Salesforce Agentforce “US$500 per 100,000 Flex Credits”	US $500	≈ €440	+19% → ≈ €524
Salesforce Agentforce “US$2 per conversation”	US $2	≈ €1.76	+19% → ≈ €2.10 per conversation
Salesforce Agentforce “US$125 / user / month”	US $125	≈ €110	+19% → ≈ €131 / user / month
Zendesk Suite Team “US$55 / agent / month”	US $55	≈ €48.40	+19% → ≈ €57.60 / agent / month
Zendesk Suite Growth “US$89 / agent / month”	US $89	≈ €78.30	+19% → ≈ €93.18 / agent / month
Intercom Essential “US$29 / seat / month”	US $29	≈ €25.50	+19% → ≈ €30.35 / seat / month

Example Monthly Cost Scenarios (Germany)

Here are estimates for what a small / larger team might pay under different vendors, using the converted / VAT-included prices above.

Scenario	Vendor / Plan	Agents or Seats	Cost per Agent (incl. VAT)	Total Estimated Monthly Cost
Small Support Team (5 agents)	Zendesk Suite Team	5	~ €57.60 / agent / month	≈ €288 / month
	Intercom Essential	5	~ €30.35 / seat / month	≈ €152 / month
	Salesforce Agentforce (user-license model)	5	~ €131 / user / month	≈ €655 / month
Larger Team / Enterprise (50 agents)	Zendesk Suite Growth	50	~ €93.18 / agent / month	≈ €4,659 / month
	Intercom higher plan / seats	50	assume ~ €80 / seat / month (for more advanced plan)	≈ €4,000 / month
	Salesforce Agentforce unmetered employee plan	50	~ €131 / user / month	≈ €6,550 / month

Please Note:

• These converted prices assume standard VAT at 19% and full pass-through; vendors may have special pricing for Germany/EU and may include or exclude VAT in their listed prices.

• Currency fluctuations mean these are estimates; actual USD→EUR rate could be slightly different at time of contract/invoice.

• Some vendors offer volume discounts, negotiated enterprise deals, or different rates for non-profits, which can substantially alter cost.

• Some features (AI, automation, conversation charges etc.) may incur additional costs not captured in per-seat licensing.

• Always check whether “seat / user” includes certain features or whether some functions are add-ons (this can greatly affect total cost).

Conclusion:

Looking above at the costing details and the drivers, we can safely surmise that knowing the cost drivers like platform licenses and token consumption and selecting the suitable use cases is the key to keep the costs optimized for an AI Agent pilot implementation.

It is also imperative to reiterate that pilot is your chance to get acquainted with the platform and its capabilities as well as finalizing your scope, use cases, users for a full-scale implementation. Your home work and preparedness will put you in the best position possible to optimize TCV (Total Contract Value) as well as LTV (Life Time Value) of your AI Agent Platform.