Agentforce and AI · for companies that already pay for Salesforce

Before you switch Agentforce on, find out what it will say. Then switch it on.

An assistant answers only as well as the records underneath it. We check the data first, launch one agent with guardrails in two weeks, and count what it actually handles. Nothing is promised that your data cannot support.

Two AI projects delivered · the engineers who built them are the ones on your call · we say no when the data is not ready

Sound familiar?

Three sentences we hear about AI.

"We turned AI on and it talks nonsense."

Weak answers are almost never a prompt problem. They are gaps in the data: empty fields, stale articles, duplicates. Fix those and the same assistant becomes useful.

"Support drowns in the same ten questions."

Status checks, how-do-I, where-is-my. An agent with guardrails takes those and hands the rest to a person, with the context attached.

"We have hundreds of clients and we grow slowly."

The expansion pipeline usually does not exist in the CRM at all. Your own base already says who should be offered what next.

What we do

Data first. One agent. Then the money in your own base.

Three steps, each priced on its own. You can stop after any of them and keep everything produced so far.

Step 1

AI readiness, inside the Health Check

$1,500 · credited in full

Which objects have gaps, which fields nobody fills, which knowledge is stale. What to fix before an assistant is allowed to answer, and what not to switch on yet. A section of every Health Check report.

  • Read-only access, five days
  • Gaps listed by object, with the fix priced
  • The credit applies to the fix
What the report contains
Step 2

Agentforce QuickStart: one agent in two weeks

$5,000 · fixed

Your first agent on one channel: knowledge questions, record lookups and a clean handoff to a person. Tone and greetings set, guardrails and grounding tested, up to ten knowledge articles prepared, analytics on from day one.

  • Three topics: knowledge Q&A, record lookup, human handoff
  • Guardrails tested before go-live, test script kept
  • Thirty-day numbers: handled, handed off, wrong
Step 3

Revenue you already own

scoped after the Health Check

A ranked list of your own accounts with a concrete next offer for each, computed from what similar clients in your base already have. Signals that fire on their own; written back into Salesforce as a task with an owner and a date. The same count with the opposite sign shows churn risk ninety days early.

  • Built on your data, not a model's guess
  • Lands in the CRM as tasks, not as a report
  • ROI counted in your numbers before you commit
How it worked at Alterra

Agentforce, Data Cloud and the Einstein features are Salesforce products and need their own licences. We tell you which ones before you buy any, and we do not resell them.

How it runs

Five days to know. Two weeks to launch. Thirty days to count.

1
Days 1–5

Readiness check

Read-only access. Gaps by object, stale knowledge, who fills what. A list of what to fix before any agent answers a customer.

2
Weeks 2–3

One agent, one channel

Topics, tone, guardrails, knowledge. Tested against the questions your team actually gets, then live for a defined group.

3
Day 30

The numbers

Handled on its own, handed off, wrong. Then the decision: more channels, more topics, or stop.

Proof

What the data said at Alterra.

Nine hundred client relationships, a Life Event object configured and empty, opportunities created only after the client had already said yes. The expansion pipeline did not exist in the CRM. Counting what the base already held found $40,100 a year of recoverable revenue and $53,000 a year of manual work.

We also say no.

At GoodLeap, Service Cloud Voice and Agentforce were discussed and scoped, and not contracted. The case says so. An agent on top of an org that is not ready is a complaint, not a feature.

Alterra AdvisorsWealth management · referrals & advocates$40,100a year of revenue their team was missing, found in their own Salesforce data

Plus $53,000 a year of manual work removed. Payback in 2.7 months.

Read the case
GoodLeapLending · mortgage0 incidentsin 13 changes to a live lending system, each one tested and reversible

Work that used to take 6–12 months now ships in weeks.

Read the case
Before you say yes

Questions about Agentforce.

Do we need extra Salesforce licences?
Yes. Agentforce is licensed by Salesforce; we tell you which licences the agent needs before you buy any, and we do not resell them.
What if our data is not ready?
Then we say so, with the list of what to fix and what it costs. The Health Check fee is credited against that fix. Switching an agent on top of bad data is the one thing we will not do.
Which channels can the first agent live in?
One channel at launch: web chat, a customer portal, Slack or email. More channels after the first thirty days of numbers.
What do "guardrails" mean in practice?
The agent answers only from the knowledge and records we point it at, says "I do not know" instead of guessing, and hands off to a person on topics you define. We test those limits before go-live and keep the test script.
Will it replace our support team?
It takes the repeat questions and the lookups. Everything else is handed to a person with the context attached. The team gets the interesting work, not less work.
How do you measure whether it works?
Thirty days after launch: how many conversations the agent closed on its own, how many it handed off, and where it was wrong. You get the numbers, not a slide.

Book a Health Check.

We reply within one business day with a kickoff slot and the access checklist. The week starts when we have access.

Fixed price. Credited in full against the next project within twelve months.