How Do You Choose a Salesforce or Agentforce Consulting Partner?

TL;DR

• Choose a Salesforce or Agentforce partner on how they work, not on how big their logo wall is: scoping discipline, data honesty, governance and a clean handoff.
• Ask every candidate for the same things in writing: a sample plan, a sample agent specification, and who on their team will actually build your work.
• The right partner leaves your team able to run and extend the agents without them. Treat the handoff plan as a selection criterion, not an afterthought.

Picking a Salesforce partner used to be a question of certifications and headcount. Agentforce changes the question. You are no longer buying configuration hours. You are hiring someone to help you put digital workers into production: agents that read your data, take actions in your org and talk to your customers. That raises the bar on judgment, governance and honesty about what is ready.

This guide gives you a vendor-neutral way to compare partners. Use it as written, and share it with your procurement team.

Start With Your Problem, Not Their Pitch

Before you talk to anyone, write one page on the outcome you want. Name the process (case triage, quote follow-up, onboarding requests), who does it today, and what “done” looks like. Note where the data lives: standard objects, custom objects, Data Cloud or an outside system.

That page does two jobs. It keeps every partner answering the same question, and it shows you quickly who listens. A partner who rewrites your problem into the product they want to sell is telling you how the engagement will go.

The Six Criteria That Matter

1. Scoping discipline

Good partners narrow scope early. Ask how they would choose the first use case. You want to hear about volume, data quality, risk of a wrong action and how easy it is to measure. Be wary of anyone who proposes a large multi-agent program before a single agent has handled real work.

2. Data honesty

Agents are only as good as the data they are grounded in. A strong partner asks about duplicates, field completeness, sharing rules and integration users before talking about prompts. If nobody asks about your data in the first conversation, expect to discover the gaps in production.

3. Agent design, not chatbot design

Ask them to describe how they define an agent’s job: its topics, its allowed actions, its guardrails and when it hands off to a person. Partners who think in terms of digital workers will talk about responsibilities and escalation. Partners who think in terms of chatbots will talk about tone and greetings.

4. Governance and security

You need clear answers on permissions, the Trust Layer, prompt injection risk, audit trails and who reviews transcripts. If an agent will call outside systems, ask how they approve and limit those tools. This is where third-party integrations need the most care.

5. Evidence you can verify

Logos and badges are easy to show. Ask for something you can check instead: a walkthrough of a working agent they built, a sample specification document, or a reference you can call. Ask whether each example came from a client project or an internal build, so you know what you are looking at. Certifications show training. They do not show judgment.

6. The handoff

Ask what your team will own on day one after go-live, and what documentation, training and runbooks come with it. A good partner plans to make themselves less necessary. If the plan assumes they stay forever to keep the agent running, factor that dependency into your decision.

A Simple Scorecard

Score each candidate from 1 to 5 on each criterion, and write one sentence of evidence beside every score.

Criterion	What a 5 looks like
Scoping discipline	Proposes one narrow, measurable first use case
Data honesty	Asks about your data before your prompts
Agent design	Describes topics, actions, guardrails and handoff points
Governance	Clear plan for permissions, logging and transcript review
Verifiable evidence	Shows real work or a reference you can call
Handoff	Written plan for what your team owns after go-live

The evidence column matters more than the numbers. If you cannot write the sentence, the score is a guess.

Red Flags

• A fixed go-live date before anyone has looked at your data.
• Promises of a full rollout across every team in the first engagement.
• No named builders, only account managers.
• Guardrails and testing described as a final phase instead of part of design.
• Results claims they cannot connect to a specific, checkable project.

Questions Worth Asking in the First Call

• What would you build first for us, and why that?
• What would make you tell us not to use an agent for this?
• Who on your team will configure and test the agent?
• How do you test edge cases and adversarial prompts before launch?
• What will our admins be able to change on their own after go-live?

How Big a Partner Do You Need?

Size is the wrong lens. What matters is whether the people on your project have done this kind of work and will stay on it. Large firms bring depth and process. Smaller teams often put senior builders directly on the work. Either can be right. Ask who is actually doing the work, and judge them.

Where Funnelists Fits

Funnelists is a Salesforce consultancy based in the Miami / Fort Lauderdale area, serving clients across the US. We build custom AI apps on Salesforce and portal templates that run on your Salesforce data, and we have built Agentforce into our SF Support Portal Template and SF Custom Portal Template. We have also built example agents we can walk you through. We would rather you use the scorecard above on us than take our word for anything.

If you are comparing partners for an Agentforce project, book an intro call to talk through your shortlist and your first use case.