Agentforce demos well. Getting it funded is a different problem. This is a practical guide to building the business case for Salesforce Agentforce: where the value actually comes from, how to size it honestly, and how to write it up so a CFO says yes.
Where Agentforce value actually comes from
Agentforce earns its keep by taking on work that is high-volume, rules-based, and currently done by expensive humans, or by making those humans faster at the judgment-heavy parts. The strongest cases are not "AI everywhere." They are one or two workflows where deflection, speed, or capacity translate directly into dollars.
- Service deflection: resolving common cases without a human, reducing cost per case and freeing agents for complex work.
- Faster resolution: surfacing the right knowledge and next action so human agents close cases faster.
- Sales capacity: handling research, summarization, and follow-up so reps spend more time selling.
- Process automation: completing multi-step back-office work that currently bounces between systems and people.
How to size it honestly
A credible business case uses your numbers, not vendor averages. Anchor it in volume and fully-loaded cost, and be conservative on the automation rate.
- Start with volume: how many cases, inquiries, or tasks per month does the target workflow handle?
- Apply a conservative automation or assist rate, then sanity-check it against a small pilot rather than a slide.
- Multiply by fully-loaded cost per interaction (salary, tooling, overhead), not just hourly wage.
- Net out the real costs: Agentforce licensing, the data and integration work to ground it, and ongoing tuning.
- Add the value that is real but harder to price: faster response, better consistency, and capacity you do not have to hire for.
The honest part most cases skip
Agentforce only performs as well as the data and workflows it sits on. If your knowledge base is stale or your CRM data is messy, budget for that groundwork in the business case, because it is the difference between an agent that helps and one that confidently gives wrong answers. Building this into the plan is what makes the case survive scrutiny, and what makes the deployment actually work.
A simple model a CFO will trust
You do not need a twenty-tab spreadsheet. You need a transparent chain of assumptions a skeptic can poke at. Walk it in this order, and show your work.
- Annual volume of the target workflow, taken from real system data, not an estimate.
- A conservative automation or assist rate, validated by a short pilot before it goes in the model.
- Fully-loaded cost per interaction, so the savings are real dollars and not just minutes.
- Gross benefit equals volume times rate times cost, then subtract licensing, grounding and integration work, and ongoing tuning to get the net.
- State the payback period plainly. A case that pays back inside a year survives almost any budget conversation.
Answering the objections before they are asked
Every AI business case meets the same three questions. Address them in the document and you remove the reasons to say no.
- "Will it give wrong answers?" Show the human-in-the-loop controls, the grounding data, and where the agent hands off to a person.
- "What about the data work?" Name it and budget for it. Pretending it is free is how these projects lose credibility.
- "Why now, and why this workflow?" Lead with the single highest-value use case and a measurable pilot, not a platform-wide vision.
Phase it so the first win funds the next
The strongest cases do not ask for everything at once. They propose one workflow, a defined pilot with a success metric, and a decision gate. When the pilot hits its number, expansion is an easy yes because the value is no longer theoretical. Structure the ask as a staged investment, and you turn a big, scary number into a series of fundable steps.
The one page that gets the yes
Distill all of it into a single page: the workflow, the volume, the conservative assumptions, the net benefit, the payback period, the controls, and the phased plan. The detailed model lives in an appendix. Decision-makers approve what they can understand at a glance and defend to their own boss, so make the summary do that work.