Sales Transformation

AI readiness for sales teams: what to fix before you buy the tools

7 mins read

A brushed silver lattice sphere resting on a stack of smoked glass plates, the bottom plate glowing red, on a deep purple background.

KEY TAKEAWAYS

  • Start with the job, not the tool. Name the task or decision the AI should improve, who uses the output and how you'll judge whether it helped.

  • AI tools work from your CRM, your documents and your permissions. If stages mean different things and fields are half filled, the tool is likely to inherit those gaps.

  • Give every data source and content library an owner and a refresh rule before an AI tool reads from it.

  • Review who can see what before rollout. Microsoft says its own Copilot works within existing permissions, so overshared content can affect its results.

  • Take a baseline and decide which actions need a person's approval before go-live. Without a baseline, it's hard to judge what the tool contributed.

Before you buy AI tools for the sales team, settle five things: a clear job for the tool, a sales process people actually follow, owners for the CRM data and content it will read, access rules you've checked, and a baseline to judge it against. That is what AI readiness for sales teams means in practice. An AI tool that works from your CRM and documents can inherit their gaps. Buying a tool is quick. Getting the foundations right takes longer, and in our view it decides much of what the tool can do.

This guide is for a sales director or CRO being offered AI tools for prospecting, call summaries, deal coaching or forecasting. It covers the checks to run before buying, the red flags, and when buying now is reasonable. If the tool in question is a forecasting product, our guide to sales forecasting software covers its input checks in more detail.

Start with the job, not the tool

Name the outcome before you look at products. Microsoft's agent design framework for its Copilot Studio product starts by asking what outcome you are trying to achieve. It lists the question not to start with: "Which tools do I need?" That's guidance for one Microsoft product, but we'd use the same order for any AI purchase.

For a sales team, write the job down in three lines:

  • The task or decision. For example, preparing account research before a first meeting, or flagging deals with no agreed next step.

  • Who uses the output. A seller, a manager in a pipeline review, or a leader in the forecast call.

  • What changes if it works. The measure, and its value today.

If the vendor's demo is the only description of the job, you aren't ready to compare tools.

AI readiness for sales teams starts with the process

AI tools read and act on your sales process as it is recorded. If stages mean different things to different teams, a tool has little to go on when telling a strong deal from a hopeful one.

Merlin5's position is that AI is only as good as the process behind it. Its sales process page names the obstacles it sees: incomplete data, inconsistent stages, weak governance and low CRM adoption. In Merlin5's view, the obstacle is rarely the technology.

Three checks before you buy:

  • Every stage has a purpose and exit evidence that sellers and managers apply the same way. Our guide to sales process stages covers how to write them.

  • Pipeline reviews and forecast calls run from the CRM, not from a parallel spreadsheet.

  • People actually use the CRM. If they don't, start with recovering CRM adoption, because a tool reading a half-empty CRM has little to work with.

Give data and content an owner

Every source an AI tool reads, whether CRM fields, call notes, playbooks or price lists, needs an owner, a refresh rule and one current version.

Microsoft's Copilot Studio framework warns that data readiness, left until late, is a frequent cause of blocked projects. Among the pitfalls it lists for agents built in that product are data that goes out of date because nobody owns or refreshes it, and knowledge sources that contradict each other.

For a sales team, that means three areas:

  • CRM data. Required fields that support a decision, a rule for duplicates and a named owner. Merlin5's commercial discipline work covers data quality and pipeline governance.

  • Sales content. One current version of each playbook, product sheet and price list, each with an owner. An assistant that finds three versions of the discount policy may not know which one is current. Our guide to building a sales playbook people will use covers the content side.

  • Call and email records. Clear rules on what is recorded, where it is stored and how long it is kept.

Check who can see what

An AI assistant that searches your systems may surface anything its users are allowed to see, depending on how it is built. Check permissions before rollout, not after.

Microsoft's security guidance for its Copilot puts it this way: "Microsoft Copilot operates within existing permissions and access controls. Overshared or poorly governed content can affect Copilot results and increase risk." That describes Microsoft's product. Other tools handle access differently, so ask each vendor how theirs works.

Before rollout, check:

  • shared folders holding pricing, contracts or customer data that are open to everyone;

  • CRM roles, including who can see which accounts, margins and notes;

  • whether the tool acts with each user's own access or with a broader system account.

Decide what needs a person, and take a baseline

Decide in advance which actions the tool can take on its own and which need a person to approve them. Our suggestion: anything that reaches a customer, changes a price or moves a forecast should have an approver.

Microsoft's framework asks teams to define what an agent can do autonomously and what needs human approval, and to treat evaluation "as part of design, not an afterthought". Among its listed pitfalls is having no baseline or benchmark.

So take the baseline before go-live: the measure from your job description, how it is collected and who reads it. If the tool is meant to help sellers sell, the four drivers of sales velocity are a sensible set of outcomes to watch. A change after rollout suggests the tool may have contributed, not that it caused the change: pricing, staffing and the market move too, so log what else changed in the same period.

Then decide where the output will be used. Merlin5's digital inside sales page describes sales technology that is bought but not embedded: the platforms sit outside daily selling, manager reviews and customer conversations, so the productivity gains don't land. Name the meeting or the moment in the seller's day where the tool's output gets used.

Red flags before you sign

  • The business case describes features, not a job and a measure.

  • Pipeline reviews still run from a spreadsheet.

  • Nobody can say who owns the CRM data or the sales content.

  • The tool needs access to every system before it has a defined job.

  • No baseline has been taken.

  • Nobody has decided who checks the output before it reaches a customer.

When buying now is reasonable

Buying now is reasonable when the job is narrow, the data it needs is in good order and a mistake is cheap to catch. An illustrative example: a tool that drafts a summary of a seller's own call notes, which that seller checks before saving. Start there, measure against the baseline, and widen the scope as the foundations are fixed.

If the bigger question is how the sales organisation should work, settle the sales operating model first. In our view, a tool bought before that decision can end up shaping it.

Where Merlin5 fits

Merlin5 treats this work as the prerequisite. Its sales process and CRM transformation work designs the stages, evidence requirements and data standards, then makes the CRM enforce them. An AI-ready foundation is one of the outcomes it lists. It is platform agnostic and has no licence to sell.

For teams building or transforming digital inside sales, Merlin5's view is that clean data, structured workflow and a repeatable motion are the preconditions for AI to work at all. Its work there includes playbooks, AI and enablement, embedded in daily selling.

Where to start

  1. Write the job for the tool in three lines: the task, who uses it and the measure.

  2. Pull ten recent opportunities and check whether stage, close date, amount and next step are filled in and believable.

  3. List the data and content the tool would read, and name an owner for each.

Frequently asked questions

What does AI readiness mean for a sales team?

It means the tool has a defined job, the sales process is followed and recorded consistently, the data and content it reads have owners, and access has been checked. It is a test of your foundations rather than of the tool.

Do we need a new CRM before we use AI?

Not as a first step. Merlin5's process FAQ says most organisations already have a platform that can support a well designed process. Fix the process, the data standards and adoption on the platform you have, then judge whether it holds you back.

Can AI fix poor CRM data?

It can help, but it can't own the fix. A tool may help spot gaps or suggest wording, but the business has to agree what your stages mean and who owns each field, and then keep the CRM current. That accountability stays with sales leadership, whatever the tool suggests.

Who should own AI tools in the sales team?

The sales leader, who is accountable for the job and the result. A named person should own running the tool day to day, including its data sources and access, so it doesn't fall to whoever is nearest when something breaks.

Sources

  1. Microsoft Learn, "Use the agent design framework", Microsoft Copilot Studio guidance, updated 23 February 2026. Product-specific guidance for Copilot Studio. https://learn.microsoft.com/en-us/microsoft-copilot-studio/guidance/agent-design-canvas-framework

  2. Microsoft Learn, "Security for Microsoft Copilot", updated 13 October 2026. Product-specific guidance for Microsoft Copilot. https://learn.microsoft.com/en-us/copilot/plan/security-microsoft-copilot

NEXT STEP

See where your sales process breaks

AI is only as good as the process behind it. Before you buy, look at your stages, data and CRM with Merlin5, starting with a sales process review.

Book A Sales Process Review

Headshot of Chris Thomas

WRITTEN BY

Chris Thomas

© 2026 Merlin5

© 2026 Merlin5