All workCase study · Code apps

ProCon Planning: a conversational CRM, built around how their team works

  • Project controls and planning
  • Netherlands
  • Power Apps code apps
  • Dataverse
  • Adoption

An overview of our partnership with ProCon Planning so far

  • 5 builds in 10 monthswe built their website with a chat agent, a LinkedIn agent MVP, their conversational CRM, a project journal app to capture project updates, and now ProCon Content for all their marketing and content creation.
  • Every build on timeeach one delivered within its estimate
  • One fixed monthly feea retainer, with output across every solution

All on one Dataverse foundation. Next: bringing AI into the ProCon methodology itself.

The challenge

Their CRM was a spreadsheet.

When we started working together, ProCon Planning didn't use AI anywhere in their work, and their leads, contacts and proposals lived in an Excel sheet. During our Copilot adoption work, the team started using Copilot Chat every day. That raised an obvious question: why can't the CRM work the same way? A standard CRM would have buried them in features they'd never use.

So we built them their own: a conversational CRM, designed around the way the team already works with AI. Grounded in good data and clear governance, it lets them manage leads, contacts and proposals simply by asking. And when the ProCon methodology goes live, every new lead will land straight in it.

Interactive demo with fictional data. Pick a prompt under the chat.
Conversational CRMASK · SEE · ACTMila de Groot

CONVERSATION

Try a prompt

Your CRM is a conversation.

Ask about your leads, pipeline, contacts or accounts — the record appears on the right.

LIVE INTERFACE

Dashboard

€605,000Open pipeline
7Open opportunities
3Leads (1 new)
41%Win rate

PIPELINE BY STAGE

Discovery1 · €45,000
Qualify2 · €110,000
Proposal2 · €205,000
Negotiate1 · €150,000
Close1 · €95,000

How we did it

From solution design to live.

  1. Solution design

    Workshops on how the team captured and followed up leads in a spreadsheet, and how they already used Copilot. Out came a solution design and a backlog.

  2. Scope agreed

    The CRM scoped and estimated up front, and delivered within ProCon's fixed monthly retainer. No timesheets.

  3. Built in sprints

    Two-week sprints: a Power Apps code app on Dataverse, with an AI agent at its heart.

  4. Adopted

    Michael tested every sprint with his own acceptance tests, and the team already knew AI from the Copilot work.

  5. Live

    Their CRM is now a conversation. Plan the day, draft a follow-up, see the pipeline.

Architecture

How it fits together

Your data stays where it is. People sign in with their Microsoft account, and the CRM reads and writes Dataverse as that person, so their existing security roles still apply. The Anthropic API only receives the request and the records it needs, never anyone's sign-in.

  • Their team, on a laptop or phone, opens the Conversational CRM, a Power Apps code app.
  • The CRM works with their Microsoft tenant as the signed-in user: Microsoft Entra ID signs people in with their Microsoft account, and Microsoft Dataverse is read and written as that person.
  • The CRM sends the Anthropic API (Claude, for the conversation) only the request.
Their teamLaptop or phone
Conversational CRMPower Apps code app
Their Microsoft tenant
Microsoft Entra IDSign-in with their Microsoft account
Microsoft DataverseRead and written as the signed-in user
Anthropic APIClaude, for the conversation
  • Blue: sign-in and data, as the signed-in user
  • Magenta: the AI request, only the request
In their words

The proof is in the pudding.

TTF is helping us turn AI from an idea into solutions our team actually uses. Every build is shaped around how we work, and the adoption support means the whole team picks it up quickly. It saves us a lot of time, every single week.
Michael DuvenageMichael DuvenageCEO, ProCon Planning B.V.
Built with
  • Power Apps code app (React)
  • Microsoft Dataverse
  • Claude (Anthropic API) for the CRM agent
  • Azure DevOps for the backlog and two-week sprints, kept up to date by Claude Code through the Azure DevOps CLI