TL;DR
- A realistic first 90 days with an Agentforce consulting partner runs in three phases: foundations (days 1-30), a narrow pilot (days 31-60) and controlled rollout (days 61-90).
- The first month is mostly about your process, your data and the agent’s job description. Very little gets built, and that is a good sign.
- By day 90 you should have one agent doing real work, measured against goals you set together, and a named owner on your team who can run it.
Your team has decided to put Agentforce to work. The next question is practical: what actually happens once a consulting partner starts? Many leaders expect an install-and-go-live story. A credible engagement is more deliberate than that, because you are putting digital workers into production, not switching on a chatbot. They read your data, take actions in your org and speak for your company.
Here is what a well-run first 90 days usually looks like, and what you should expect to receive at each step. Timelines vary with scope and data. Treat the phases as the shape, not a contract.
Days 1-30: Foundations
The first month is mostly listening and design. Expect your partner to spend more time with your people and your data than with Agent Builder.
What happens:
- Workshops to map repetitive, high-volume work that is worth handing to an agent.
- A review of the data the agent will touch: duplicates, missing fields, sharing rules, consent and integration users.
- Choosing the first use case by impact, data quality and the risk of a wrong action.
- Designing the agent’s job: its topics and actions, guardrails and the points where it hands off to a person.
- Sandbox setup and a permission model for the agent.
What you should receive:
- A short, ranked list of use cases with the reasoning for the first pick.
- A data gap report with the fixes needed before launch.
- An agent specification: role, instructions, topics, actions and escalation rules.
- Success measures you agree on together, written down before anything is built.
This phase prevents the most expensive mistake in agent projects: building on data that is not ready.
Days 31-60: A Narrow Pilot
With the design agreed, building starts. The goal is not a sprawling multi-agent system. It is one focused agent that proves the approach works in your real org.
What happens:
- Configuring the agent: topics, instructions, standard actions and any custom actions in Flow or Apex.
- Grounding work, such as knowledge article cleanup, retrieval scope and any Data Cloud connections.
- Building integrations for actions that reach outside Salesforce.
- Testing normal cases, edge cases and attempts to misuse the agent, including prompt injection.
- A pilot with a small, friendly group of users on one channel.
What you should have at day 60: A live agent handling real requests, with results tracked against the measures from month one, clear escalation paths and full logging. You also get a go or no-go decision based on what the pilot showed, not on how the agent looked in a walkthrough.
Days 61-90: Controlled Rollout
The last month moves from “does it work?” to “how do we run it safely as usage grows?”
What happens:
- A phased rollout to more users or more traffic, widening as results hold.
- Training and change management for the teams working alongside the agent.
- An operating model: who reviews transcripts, who tunes instructions and who watches usage and consumption.
- Dashboards and alerts for the agent’s performance and escalations.
- A plan for the next use case or channel.
What you should have at day 90: One agent in production doing real work, a repeatable process for adding the next one, and a named owner on your team who can run it without the partner in the room.
Checkpoints at a Glance
| Day | You should have | You should not have |
|---|---|---|
| 30 | A written plan, agent spec and data fix list | A half-built agent on unreviewed data |
| 60 | A live pilot with tracked results | A wide launch with no measures |
| 90 | A production agent, an owner and an operating model | A dependency on the partner to keep it running |
Red Flags in the First 90 Days
Watch for a partner who:
- Wants to start building on day one, before anyone has looked at the data.
- Promises a full rollout across every team inside the first 60 days.
- Treats guardrails and testing as a final step.
- Cannot walk you through an agent they have built.
- Has no plan for what your team will own after go-live.
A credible partner pushes back on scope in month one and keeps the pilot narrow in month two. That pushback protects you.
What Your Team Brings
The partner cannot do this alone. Plan for:
- An executive sponsor who can make scope calls quickly.
- A process owner who knows how the work is done today.
- An admin who will own the agent after go-live.
- Access to sandboxes, data and the people who use the process.
The Bottom Line
The first 90 days are not about launching the most agents. They are about proving that one digital worker can do real work safely inside your Salesforce org, with your data and under your rules. Get the foundations right, keep the pilot narrow and only then widen the rollout.
Where Funnelists Fits
Funnelists builds custom AI apps on Salesforce, from the Miami / Fort Lauderdale area, serving clients across the US. We have built Agentforce into our SF Support Portal Template and SF Custom Portal Template, along with example agents. We would plan a first engagement in the phases above, starting with your data and a single, measurable use case.
If you are planning your first Agentforce engagement, book an intro call to talk through your first 90 days.
Agentforce & AI Agents
Autonomous agents that predict, automate, and execute revenue-driving actions across sales, service, and marketing.
Read Next
Agentforce ROI: What Actually Moves the Needle for Busy Leaders
Agentforce has moved past pilots. Here's a simple way to prove ROI: pick one process, capture a baseline, and track leading and lagging metrics.
How Do You Choose a Salesforce or Agentforce Consulting Partner?
A vendor-neutral way to compare Salesforce and Agentforce partners: six criteria, a simple scorecard and the red flags to watch for.
Why Agentforce Third-Party Integrations Demand Custom Guardrails
Salesforce admins and solution architects integrating Agentforce with external tools via MCP face supply-chain and privilege risks that native guardrails do not fully cover. This post shows how to add targeted custom controls without heavy custom code.
