Pre-sales scoping for Agentforce vs. traditional Salesforce
The core difference: scope that moves
In a traditional Salesforce implementation, the pre-sales scoping conversation is largely about configuration boundaries. You map business processes to standard objects, identify gaps, agree on customization depth, and produce an estimate. The unknowns are real but bounded — you are fitting a known platform to a known workflow.
An Agentforce or agentic-AI engagement introduces a different kind of unknown: the agent's decision surface. An agent does not just execute a defined sequence of steps; it interprets context, selects actions, and can branch in ways that a conventional workflow map does not capture. That changes what you need to clarify before you can price the work.
What the scoping conversation has to cover
For a traditional Salesforce project, a solid scoping session covers:
- Which clouds and standard features are in scope
- Data model changes and integration points
- User roles and permission sets
- Migration complexity and volume
- Go-live and training requirements
For an Agentforce engagement, those questions remain, but you need to add a second layer:
- **Agent scope and authority** — what decisions is the agent allowed to make autonomously, and where must it hand off to a human?
- **Action inventory** — which Salesforce actions, external API calls, or data sources will the agent be able to invoke?
- **Guardrail design** — what topic restrictions, escalation paths, and audit requirements apply?
- **Grounding data** — what knowledge articles, data cloud segments, or retrieval sources will the agent reason over, and how current does that data need to be?
- **Failure and fallback behavior** — what happens when the agent cannot resolve a request with sufficient confidence?
Each of these adds estimation surface that simply does not exist in a conventional CRM build. Skipping them in pre-sales means the estimate delivered to the client is almost certainly wrong.
Why the estimate is harder to anchor
In a standard implementation, story points or day-rate estimates map reasonably well to feature lists. You have done similar projects before, and the variance is manageable.
With agentic AI, the effort is more sensitive to decisions that are often still open at the pre-sales stage: which large language model the org will use, how many agent topics are in scope, whether the team has prior Agentforce certification, and how mature the client's data hygiene is. A one-topic service agent and a multi-topic sales-assist agent with external integrations can differ by an order of magnitude in delivery complexity, even if they look similar in a brief kick-off call.
This means your pre-sales artifact needs to surface those open decisions explicitly rather than papering over them with a wide range estimate.
What a structured pre-sales output should include
A useful pre-sales package for either engagement type covers a plain-language brief, a proposed solution narrative, a technical specification, and a pricing reference. For an Agentforce engagement specifically, the technical specification needs to document the agent's action inventory and authority boundaries — not just the org configuration — and the estimate needs to call out the assumptions that could move the number significantly.
The more senior the reviewer on the client side, the more they will want to see the reasoning behind the estimate, not just a bottom-line number. An architecture review that addresses agentic-specific risks — data grounding quality, prompt injection surface, human-in-the-loop design — gives a consulting proposal considerably more credibility than a slide deck with a price on the last page.
Practical steps for consultants
1. Build a standard intake checklist that includes the agentic-specific questions above.
2. Produce the technical spec and estimate as separate documents so clients can review scope before they react to price.
3. Flag open assumptions in writing. If data grounding quality is unknown, say so and quantify the risk in the estimate.
4. Plan for a scoping revision cycle. Agentic engagements frequently surface new requirements once the client sees a working prototype.
Not sure if you even need an agent?
Before you spend an hour on a full scoping conversation, run KevTech's free [Agentforce scoping wizard](/self-serve/agentforce-scope/new) — describe the business in a few fields and get an honest evaluation of whether an agent is warranted, plus a full technical spec per agent if one is.
Get your first scoped output at no cost
KevTech Solution Engine generates a Brief, Solution, Technical Spec, Estimate, and Pricing reference from a single intake. Every account receives one free pipeline run — results are shown redacted until you unlock, starting at $97 for clients or $197 for independent consultants. Start your first scope at [/self-serve/signup](/self-serve/signup).