AI and CRM as a revenue engine
From customer signals to explainable action in Dynamics 365.
Your organisation already knows a lot about its customers, but the information sits in different places.
We connect the relevant signals across CRM, ERP, Microsoft 365 and other systems. AI helps prioritise the next action and explain why, while people remain in control of important decisions.
Book an introductory callIs this a fit for you?
The approach works best for larger organisations with a customer journey that repeats often, data in several systems and a commercial result that can be measured. If the data or the process is not ready, the first recommendation may be to fix the foundations before AI is introduced.
Three layers
Data and signals
Relevant customer information is made available across CRM, ERP, Microsoft 365 and other sources. Fabric can be used as an analytical layer when the need is there.
- CRM
- ERP
- Microsoft 365
- Customer service
- Other sources
- Fabric when needed
AI in the workflow
Agents gather context and suggest or carry out the next action within clear mandates.
- Gathers context
- Suggests or carries out the next action
Within clear mandates. People remain in control of important decisions.
Where people work: Dynamics 365, Outlook and Teams
Measurement and decision support
Management sees the effect on conversion, response time, margin and customer loyalty after operating costs have been deducted.
- Conversion
- Response time
- Margin
- Customer loyalty
- Net of operating costs
What it looks like when it works
A car dealer has signals everywhere: web leads, test drives, trade-ins, financing, leases coming to an end, service visits and workshop history. Most of them need no action. The solution should find the few that do and leave the rest.
An example: a customer’s lease ends in four months, and the car is booked into the workshop next week. The salesperson gets a suggestion in Dynamics 365 to get in touch before the service visit. The reasons are shown alongside it: the end date of the lease, the mileage from the last service and an earlier enquiry about electric cars. The salesperson decides whether and how to contact the customer.
The same goes for new web leads. They are routed to the right salesperson as they come in, and the sales manager can see which ones have not been followed up.
What should be measured?
The effect is what happens on top of what would have happened anyway. Before we start, we agree on three things:
- The metric to move, such as quote-to-order conversion, response time on new leads or the share of customers who renew.
- The comparison: a control group that works as before, or an agreed baseline from historical data.
- The operating cost: licences, AI consumption and management, so the gain is calculated net.
A worked example: a company sends 500 quotes a year averaging NOK 500,000. If conversion rises by two percentage points compared with the control group, that could mean NOK 5 million in additional sales before operating costs. If the company has NOK 200 million in recurring revenue from service agreements, subscriptions or memberships, one percentage point lower churn retains NOK 2 million a year.
Why it fails
Data
Duplicates, outdated contacts and fields nobody owns lead to poor suggestions. The data has to be in order in the process being supported.
Mandate
An agent needs a defined task, an owner and clear rules on what it can do on its own and what a person must approve. Without that, it is either ignored or does things nobody asked for.
Privacy
The fact that data exists does not mean it can be used for anything. Purpose, legal basis and access must be settled before customer data is used in new processes.
Fragmented systems
A chatbot on top of a CRM that is not connected to ERP, customer service and email sees only part of the customer. Signals have to be connected where they arise, with standard integrations and Cartagena Link to ERP.
How we start
We start by mapping one well-defined revenue process. The mapping normally takes two weeks, at a fixed price agreed before we begin.
Together with the business and IT, we go through:
- where opportunities, customers or margin are lost today
- which signals are relevant, and which should be ignored
- which action the system should suggest or carry out
- what a person should still decide
- which data, integrations and access need to be in place
- how the additional effect and the operating cost will be measured
You get a concrete basis for decision with the target process, signals, data needs, governance principles, a measurement plan and an estimate for the first version.
The mapping does not include development or changes to the production environment. After the review, you decide whether to go further.
Book an introductory callQuestions and answers
Do we need to be on Dynamics 365 already?
No, but the solution is built on Dynamics 365 and Dataverse. If you use a different CRM today, the mapping shows whether the move should come before or together with the first version. If you use Microsoft 365, many of the signals are already in Outlook and Teams.
Do we need Microsoft Fabric?
Not necessarily. Many processes can be supported with data that is already in Dataverse and Microsoft 365. Fabric can be the right choice when you need to combine large volumes of data from several systems, calculate things like churn risk or purchase likelihood, or give management analysis across systems. The mapping shows whether the need is there.
Do the agents have to be allowed to contact customers automatically?
No. You decide what an agent can do on its own and what a person must approve. Many start with the agent only suggesting the next action and showing the reasons, while the salesperson decides. The mandate can be widened step by step once the suggestions have proven reliable. Actions are logged, so it is traceable what the agent did and why.
How do we measure whether the investment pays off?
Before we start, we agree on a metric, such as quote-to-order conversion, response time or the share of customers who renew. The effect is measured against a control group that works as before, or against an agreed baseline from historical data. Licences, AI consumption and management are then deducted. That shows you the net gain, so you can decide whether to expand the solution.
Why Cartagena
Microsoft provides the platform and the general building blocks. The business logic should be yours. We help you turn it into data models, criteria, agents and workflows that can be measured and developed further.
Cartagena has delivered solutions on the Microsoft platform since 2006. The senior specialists who design the solution also help build and manage it.
Book an introductory call