SSalesCrew

Is your CRM ready for AI agents?

Answer 12 questions in about three minutes. Get a score out of 36, a score per area and the first gap to close.

An admin approves every new account by hand. Nothing is created until then. We reply by email; no newsletter, no sequence.

  • Full rubric published
  • Runs in your browser
  • No email required
app.salescrew.io/data-bank
The contact bank with segments and suppression

Score your CRM: 12 questions

0 = not true. 1 = true in places. 2 = mostly true. 3 = true and you could show it today.

  1. Data
    1. Are your contacts, companies and deals in one CRM today?
    A 3 looks like: One system of record. Three exports from three tools scores 0
  2. Data
    2. Is the data clean enough for an agent to act on?
    A 3 looks like: Duplicates merged, addresses mostly valid, junk already removed
  3. Data
    3. Are deal stages dated events, not feelings?
    A 3 looks like: Every stage change has a date and a named trigger
  4. Access
    4. Can an AI client read and write the CRM through an API or MCP server?
    A 3 looks like: An MCP server or API with write scopes, working today
  5. Access
    5. Are agent tokens scoped per user and area?
    A 3 looks like: Each token sees only what its user may see, and you can revoke it
  6. Access
    6. Can an agent do everything a person can in the app?
    A 3 looks like: Every screen action has a tool twin with the same effect
  7. Control
    7. Do external sends wait for a person's approval?
    A 3 looks like: Emails and messages an agent writes land in one approval queue first
  8. Control
    8. Is every agent action written to an audit log?
    A 3 looks like: Who or what acted, which tool, and when, for users, tokens and agents
  9. Control
    9. Can you switch one agent, or all of them, off at once?
    A 3 looks like: Per-agent modes plus one kill switch you own
  10. Ownership
    10. Does one named person own the CRM and its agents?
    A 3 looks like: A named person with time in their calendar for it
  11. Ownership
    11. Is your do-not-contact list enforced by the system?
    A 3 looks like: Suppression is checked automatically before anything sends
  12. Ownership
    12. Is AI spend capped?
    A 3 looks like: An alert before the cap and a hard stop that pauses AI only
0 of 12 answered
0 / 36: Build the base first

Get the data into one place and name an owner. Those two steps raise the score faster than any tool.

Data
0 / 9
Access
0 / 9
Control
0 / 9
Ownership
0 / 9

The short answer

  • A CRM is ready for AI agents when four areas are in place: clean data, agent access, controls on what agents do, and a named owner.
  • The scorer asks 12 questions worth 0 to 3 points each, 36 in all, with 9 points per area.
  • 27 or more means ready: start agents in review mode. 15 to 26 means fix your 0s and 1s first. Under 15 means data and ownership come first.
  • The weakest area is where to start, because an agent is only as safe as its least controlled step.

How the score works

Score each question 0, 1, 2 or 3. Add all 12 for the total, and add each group of three for the area score. Every question weighs the same.

  • 27 to 36, ready for agents. Start each agent in review mode and move one action to auto after a clean week.
  • 15 to 26, ready once you close the gaps. Fix every 0 and 1, weakest area first.
  • 0 to 14, build the base first. Move the data into one CRM and name an owner.

When two areas tie for weakest, the tool names the first in this order: Data, Access, Control, Ownership.

The 12 questions and what a 3 looks like

Three questions per area, three points each.

#AreaQuestionWhat a 3 looks like
1DataAre your contacts, companies and deals in one CRM today?One system of record. Three exports from three tools scores 0
2DataIs the data clean enough for an agent to act on?Duplicates merged, addresses mostly valid, junk already removed
3DataAre deal stages dated events, not feelings?Every stage change has a date and a named trigger
4AccessCan an AI client read and write the CRM through an API or MCP server?An MCP server or API with write scopes, working today
5AccessAre agent tokens scoped per user and area?Each token sees only what its user may see, and you can revoke it
6AccessCan an agent do everything a person can in the app?Every screen action has a tool twin with the same effect
7ControlDo external sends wait for a person's approval?Emails and messages an agent writes land in one approval queue first
8ControlIs every agent action written to an audit log?Who or what acted, which tool, and when, for users, tokens and agents
9ControlCan you switch one agent, or all of them, off at once?Per-agent modes plus one kill switch you own
10OwnershipDoes one named person own the CRM and its agents?A named person with time in their calendar for it
11OwnershipIs your do-not-contact list enforced by the system?Suppression is checked automatically before anything sends
12OwnershipIs AI spend capped?An alert before the cap and a hard stop that pauses AI only

A self-assessment rubric. Every rule and cut-off is on this page.

Three worked scores

All 3s scores 36: ready for agents. Start them in review mode anyway, because trust in an agent is earned one approved draft at a time.

All 1s scores 12: build the base first. Every area sits at 3 of 9, so the tool names Data as the place to start.

A mixed team scores 18. Data 8 (3, 3, 2), Access 2 (1, 0, 1), Control 1 (0, 0, 1) and Ownership 7 (3, 2, 2). That's ready once you close the gaps, and Control is the weakest area. Add an approval queue and an audit log before any agent sends.

Access and Control, built into SalesCrew

Every action in SalesCrew is also an MCP tool with the same effect, on per-user tokens scoped by channel, profile and area. External sends wait in one approval queue, every tool call is written to the audit log, and a kill switch stops every agent at once.

Questions

What makes a CRM ready for AI agents?
Four things: clean data in one place, agent access through an API or MCP server with scoped tokens, controls such as an approval queue and an audit log, and a named owner. This scorer gives each area 9 of the 36 points.
What score counts as ready?
27 or more out of 36. From 15 to 26, fix the questions you scored 0 or 1 first. Below 15, start with the data and a named owner before you add agents.
Should AI agents send emails without approval?
Start every agent in review mode, where it drafts and a person approves each external send. In SalesCrew, every agent ships in review, and external sends wait in one approval queue by default.
Do I need an MCP server to use AI agents with my CRM?
You need a way for the agent to read and write, and MCP is the standard way for Claude, ChatGPT and Cursor to connect. SalesCrew's MCP server has 140+ tools, one for every action in the app.

Related

Get Access and Control in place on day one

SalesCrew ships with 140+ MCP tools, one approval queue and a full audit log. Unlimited seats from $299 a month.

An admin approves every new account by hand. Nothing is created until then. We reply by email; no newsletter, no sequence.

Want to see it first? Book a demo