Cognism's MCP lets Claude and ChatGPT pull verified GTM data straight into a conversation, without a dashboard or a CSV in sight. Below are 20 real prompts GTM teams are already running: prospecting, TAM and territory planning, account research, pipeline analysis, marketing operations, and multi-tool workflows across Salesforce, HubSpot, Slack, Gong, and Dreamdata. It's read-only, so nothing here changes a record without you doing that yourself.
If you've already read our article on "What is MCP in AI, and why does it matter for GTM stacks?". This piece skips that groundwork and gets straight to the point: 20 real MCP use cases, organised by the work GTM teams actually do.
Note: Cognism's MCP is read-only. These examples illustrate how information can be surfaced across tools, not how records are automatically updated.
One of the biggest time-savers for SDRs is account research. Many of the prompts below follow that same pattern: understanding an account, validating fit, and finding the right contacts without jumping between multiple tools.
"I mainly use the MCP to speed up account research. I can quickly check company headcount, see how big specific teams like Sales and Marketing are, to work out if the account is a good fit, and then ask it to find the right people based on the job titles I normally target.
Usually, I'd be jumping between Cognism, LinkedIn, and doing a lot of that manually, so having it all in one workflow was definitely useful."
Shiwam Singh, SDR at Cognism
With Cognism MCP enabled, these kinds of tasks can happen inside a single conversation. The prompts below are grouped by workflow, from prospecting and account research through to governance and compliance. Choose one that mirrors a process you already do manually and see what comes back.
Finding the right accounts and contacts is the most common starting point, and it's where a compliance check earns its place inside the prompt itself.
"Using this list of 40 company domains, pull headcount, sales-team headcount, and verified contacts for key sales decision-makers at each company."
Top tip: Matching by domain instead of name prevents confusion between distinct companies that may share a generic name. It also ensures clarity throughout the entire list, eliminating the need for individual lookups.
"Find 40 target accounts in the Nordics by buying-signal strength, and tell me which three have hired a new VP of Sales in the last quarter."
Top tip: This is an example of account prioritisation, together with justifications, built on real buying signals rather than a sorted list alone. It can be done at scale, too: at Cognism, our SDRs have run hundreds of these lookups against a full target list in a single session. This work would previously have meant a manual CSV export.
"Build a list of SaaS companies in the UK with 50 to 200 employees that have posted a RevOps or sales ops role in the last month, and give me the hiring manager's verified contact details."
Top tip: If you're curious as to whether a team is gearing up for a change in its operations, consider using this prompt to uncover the hiring signals at play.
"Here's a list of 25 LinkedIn profile URLs from a conference attendee list. Match each one to a verified contact record, and flag anyone you can't confidently match."
Top tip: This prompt is handy after a conference, when resolving profiles one at a time is one of the more tedious parts of working the list. This turns it into a single batch request.
Sizing a market and splitting it fairly used to mean a spreadsheet and a lot of manual filtering. A single prompt does both.
"Size the addressable market for mid-market manufacturing companies across the UK, France, and Germany with 100 to 500 employees, and break down how many verified contacts we have at each."
Top tip: A natural-language TAM analysis like this stays current every time you ask it – it doesn’t go stale the way a one-off export does.
"Split that same market into three even territories by company count, and suggest which regions each territory should be based around."
Top tip: Territory planning like this starts from actual account density, so you no longer have to guess where the opportunity lies.
"Within our top 50 existing accounts, show me which departments or subsidiaries we haven't engaged yet, and pull verified contact details for relevant decision-makers in those areas."
Top tip: Whitespace analysis usually means a periodic report. Run it as a live query instead, as it will be accurate as of the day you ask your AI assistant.
Research before a call is where manual lookups cost the most time relative to the value they add, since the same three or four checks happen before almost every meeting.
"I've got a call with [target account] tomorrow. Pull their recent leadership changes, company size, and any buying signals from the last 30 days."
Top tip: One prompt replaces three separate lookups, and the leadership-change check catches the stakeholder who left before anyone finds out mid-call.
"Cross-check the attendee list for tomorrow's call against our CRM, and tell me if anyone's changed job title since we last spoke to them."
Top tip: It’s worth doing this preparation before a multi-stakeholder meeting, as multithreading a deal only works if the people you’re tracking remain in their designated roles.
"Pull the EU sales headcount for [target account], and tell me whether that supports the deal size we've assumed in the pipeline."
Top tip: This serves as a quick reality check before a rep spends a week on an account that may turn out to be smaller than the suggested opportunity size.
A RevOps lead's questions can end up repeating week over week. These are the ones that can be helpful to turn into a standing prompt.
"Which deals in my pipeline haven't had contact activity in the last three weeks, and are any of them tied to a stakeholder who's since left the company?"
Top tip: This surfaces stalled deals and the reason they've stalled in the same answer, without a separate pipeline management review.
"Review all active opportunities forecast to close this quarter and identify anywhere the primary contact has changed role, left the company, or has limited contact coverage. For each account, suggest alternative stakeholders with verified contact details who could help keep the opportunity moving."
Top tip: Forecasts often assume the right stakeholders are still engaged. This prompt helps uncover deal risk from contact changes before it shows up in your numbers, while giving reps a clear path to reestablish momentum.
Reviewing inbound leads is repetitive by nature, which makes it a good fit for a prompt that does the same check every time.
"Find French companies currently showing intent around AI sales tools and GTM automation. Rank them by intent strength and ICP fit, then provide verified contacts for senior RevOps, Sales Operations and GTM leaders to invite to our upcoming webinar."
Top tip: Rather than inviting every company that matches your ICP, use intent data to focus your event promotion on accounts already researching the topics you'll discuss. This helps marketing teams prioritise outreach to companies more likely to engage with the content.
"This inbound lead just filled out our demo form. Check the company size, industry, and whether the email domain matches a real company before I route it to a rep."
Top tip: Catching a personal or mismatched email domain before it reaches a rep saves a wasted follow-up.
"Look at the last 20 inbound leads from our Germany campaign, and tell me which ones meet our ICP criteria for company size and industry before I mark them as MQLs."
Top tip: Batch lead scoring like this beats working through a list one lead at a time.
This is where Cognism's MCP shows its actual value: one conversation, several tools, and zero tab-switching in between.
"Pull the verified contact details for everyone on this list, add the ones with a valid mobile number to our HubSpot sequence, and post a summary in our outbound Slack channel."
Top tip: Use this one request, connecting three different systems, to streamline your process.
"Check this account's intent signals in Cognism, then look up whether they've shown any activity in our Dreamdata attribution data in the last 30 days."
Top tip: This is a combined read across two data sources that would otherwise mean two separate logins.
"Check the notes from our last Gong call with [target account], then pull their current contact details in case anyone's moved role since that call."
Top tip: Use this prompt to follow up with the right context, on the right person, without switching between two separate tools first.
Before sharing a list, it's important to perform a couple of checks independently. Keep in mind that each check works because Cognism's MCP can look something up, not because it can modify or act on that information. This one-way structure means there is no write-back path for a security review to examine.
"Before this list goes to our German outbound team, confirm every contact is opt-out clear, and that any Do Not Call or TPS flags are actually being masked, not just recorded."
Top tip: This should be run as an occasional spot check, although Cognism will automatically flag DNCs in results returned by its MCP.
"Based on what you know about my role, my company, and how I work, suggest three practical ways I could use Cognism's MCP day to day."
Top tip: This is a good move once you've tried a few of the prompts above: instead of working through someone else's list, ask it to build your own.
The last prompt in this list is typically the best place to start: inquire about your own role and see what it recommends first. Don't hesitate to play around with the Cognism MCP, as experimentation can often lead to surprising insights.
If you're feeling stuck on how to structure or frame a prompt, reach out to your AI assistant, such as ChatGPT or Claude. It can provide valuable guidance and suggestions to help you articulate your thoughts more clearly. Remember, the key is to explore and adapt until you find the best way to communicate your needs.
Ready to bring trusted B2B data into your AI workflows? Explore what Cognism MCP can do with your own data.