Microsoft Copilot or a Custom AI Solution?

Every organization with Microsoft 365 eventually asks itself the question: is it enough to switch on Copilot, or should you invest in a custom AI solution? It’s the right question to ask – both cost real money, but they solve different problems. Here’s the math, the boundaries, and the division of labor that makes it add up. And since both items tend to land in the same AI budget, they need to be compared against the same yardstick.

What the Copilot License Gives You

Microsoft 365 Copilot puts AI support into the tools your staff already use: drafts in Word, analysis in Excel, summaries in Outlook and Teams. Rollout is fast, no custom development is required, and the benefit spreads thin but wide – a little time saved for many people, every day.

The pricing model is just as simple: one license per user per month, with a list price that has hovered around USD 30. In practice, that works out to roughly SEK 4,000 per employee per year, depending on exchange rate and contract terms.

The benefit is real but hard to measure: minutes here and there rather than whole services, and it depends on how clean and permission-controlled your data is within the Microsoft environment. A pilot with a limited group and simple before-and-after measurements gives you your own number instead of the vendor’s.

What Copilot Can’t Reach

Copilot’s strength – that it lives in everyday Microsoft work – is also its limit. Three areas fall outside it:

• Your own systems. The ERP system, the production database, and the custom-built ticketing system aren’t reachable without customization work via connectors – and at that point, you’re effectively in a development project anyway.
• Industry-specific workflows. The pricing calculation, risk assessment, or quality control that makes up your core business requires logic, data, and integrations a general office assistant doesn’t have.
• Customer-facing services. Copilot helps your employees – it never becomes an AI feature in your app, your customer portal, or your website.

That’s usually where the big leverage lives: high-volume workflows, a clear cost per case, and direct customer impact. A good litmus test: if the need is described with the words “in our system” or “toward our customers,” it’s probably outside the license’s reach.

The Math: The Breakpoint

The license cost looks small per user but grows linearly with the organization – and that’s where the comparison gets interesting:

Compare that to a custom AI solution: a narrowly scoped first solution often starts around a couple hundred thousand SEK, while larger initiatives with multiple integrations run up toward a couple million – plus ongoing maintenance. What drives the cost is integrations, data quality, and security requirements.

The worked example that shows the breakpoint: an organization with 200 employees pays roughly SEK 800,000 per year for Copilot – every year. For a comparable one-time amount, a specific high-volume workflow – case handling or quoting, for example – can be automated for real. If that workflow costs two full-time positions today, the payback period is short, and the effect is lasting. The point isn’t to skip Copilot – it’s to compare the benefit per krona, not just the price. At the same time, budget for maintaining the custom solution – models get updated, integrations change, and users want more. A reasonable rule of thumb is to budget an ongoing annual share of the investment, just as you would for any other business-critical software.

The Answer Is Often Both

Most organizations land in the right place with a clear division of labor: Copilot for general office productivity, a custom solution for the core workflows where specific automation delivers the most leverage. What matters is that the decision is made deliberately – licenses for everyone out of habit and zero investment in core workflows is the most common form of AI budget that delivers no payoff. The split also makes follow-up more honest: license benefit is measured in broad productivity, the custom solution in concrete workflow numbers – and each can be reassessed on its own.

Start by listing your three most expensive manual workflows and running the numbers on them. At Weapp, we help companies do exactly that analysis and build AI solutions where they pay off best – get in touch and we’ll look at your math together.