An AX 2012 to Dynamics 365 Finance & Operations upgrade is not priced from the number of users. It is priced from eight things: how much custom code you carry, how many integrations you run, how much history you want to keep, how clean your data is, how many legal entities and countries you operate, how much of the process you want to change on the way, how you will test, and who owns the work after go-live. Two companies with the same user count can land on figures three or four times apart. This article walks through each driver, the ratios we see between them on real projects, the lines that are usually missing from the first budget, and how to get to a number you can defend in front of a board.
Why the first quote is usually wrong
Most upgrade budgets start with a partner’s day-rate multiplied by a guessed duration, plus the licence. The guess is made before anyone has run the code and data assessment, so it is anchored on the last project the partner did, not on your system. The second mistake is treating the upgrade as a technical move of the same solution. It is not: X++ customisations are rewritten as extensions, the client becomes a browser, reporting moves to Power BI and electronic reporting, and integrations move to data entities and Azure. Each of those is a decision with a cost, and the first quote has not made any of them yet.
The honest answer to “what does it cost” before an assessment is a range with a wide spread. The honest answer after an assessment is a figure with a ten to fifteen percent contingency. The assessment itself is the cheapest money in the whole project, which is why we start there.
The eight cost drivers
1. Custom code (X++) volume and quality
The largest single driver on nearly every project. AX 2012 code is counted in objects and in the models they live in; the assessment tool reports how many are overlayered (modify standard code) versus extended. Overlayering is not allowed in Dynamics 365, so every overlayered object is either dropped, rewritten as an extension, or replaced by standard functionality that did not exist in 2012. In our experience a third to a half of customisations on a ten-year-old AX system are no longer needed, because the standard product caught up or the business process they served has gone. Finding that out costs a workshop per module; not finding it out costs a rewrite nobody uses.
2. Integrations
Count every file drop, AIF service, direct SQL read and scheduled job that moves data in or out: banks, carriers, web shops, payroll, warehouse systems, EDI, reporting databases. Each becomes a data entity, a recurring integration, a Logic App or a Power Automate flow. A system with three integrations and a system with thirty are different projects. Direct SQL reads by other systems are the ones most often forgotten, because nobody in the ERP team built them.
3. How much history you keep
Opening balances only, two years of transactions, or everything since 2012. The first is cheap and clean, the last is the most expensive line in the data workstream and the one most often demanded by finance and audit. There is a middle path: migrate balances and open items, keep full history in a read-only reporting copy of AX. We covered the mechanics in keeping historical balances intact.
4. Data quality
Duplicate customers, vendors with no payment terms, items with dead dimensions, charts of accounts that grew by accretion. Every one of those is fixed either before migration (cheap, by your team, in AX) or during it (expensive, by the project, under deadline). Budget the clean-up as its own line and start it before the partner is on site.
5. Legal entities, countries and localisations
Each legal entity adds configuration, each country adds tax and statutory reporting, and each localisation that was custom-built in AX 2012 has to be checked against what Microsoft now ships. Multi-country projects also add test cycles, because the same change has to be proven in each jurisdiction.
6. Process change
A “like-for-like” upgrade is the cheapest on paper and the most disappointing in practice: the business pays for a new system and gets the 2012 process in a browser. Every process you decide to improve on the way (approvals, warehouse mobile, bank reconciliation, budgeting) adds design and training effort and removes custom code. The right amount of change is a judgement; the wrong amount is “we will decide later”.
7. Testing
Microsoft now ships updates to Dynamics 365 several times a year, so testing is not a one-off go-live cost. The upgrade project has to leave behind a regression pack (scripted or automated) that the business can rerun. Projects that skip this line pay for it every update wave. See testing update waves without stopping the business.
8. Who owns it afterwards
Hypercare, the first month-end on the new system, the first year-end, and the support model after that. Internal team, partner retainer, or a mix. This is the line most often left out of the business case and the one that decides whether the project is judged a success a year later.
The ratios we see between the lines
Figures differ by company; the proportions are surprisingly stable. On a typical mid-sized AX 2012 upgrade with moderate customisation, the effort splits roughly like this:
| Workstream | Share of project effort | What moves it up |
|---|---|---|
| Assessment, design and fit-gap | 10 to 15% | Many legal entities, undocumented processes |
| Code: drop, rewrite as extensions, replace with standard | 25 to 35% | Overlayered objects, custom modules, no original developers |
| Data migration and clean-up | 15 to 20% | Full transaction history, poor master data |
| Integrations | 10 to 20% | Count of external systems, direct SQL readers |
| Testing and training | 15 to 20% | Countries, user count, no existing test scripts |
| Cut-over, hypercare and first close | 5 to 10% | Short cut-over window, no dress rehearsal |
Two things follow from the table. First, code and data together are usually more than half of the effort, and both are largely under your control before the project starts. Second, the licence, which is where most budget conversations begin, is not in the table at all: it is a separate recurring cost, and on most projects the first-year services figure is a multiple of it.
The lines missing from the first budget
- Azure and environments. Dynamics 365 runs in Microsoft’s cloud, but development, test and training environments, plus Azure services for integrations, are billed as you use them during the project.
- Reporting. SSRS reports and Management Reporter definitions do not all carry across. Rebuilding financial reporting in electronic reporting or Power BI is a workstream, not a task.
- Third-party ISV solutions. Every add-on you run on AX 2012 either has a Dynamics 365 version (new licence, new install) or it does not (replace, or rebuild).
- Your own people’s time. Process owners in workshops, key users in testing, finance in reconciliation. Usually a third or more of the total effort, and almost never costed.
- Running AX in parallel. Hardware, licences and support for the old system until it is decommissioned, often six to twelve months after go-live for reference and audit.
- Contingency. Ten to fifteen percent after an assessment; more before one. A budget with no contingency is a forecast of a change request.
How to get to a number you can defend
- Run the code and data assessment first. Microsoft’s upgrade analysis tools plus a module-by-module review with the people who use the system. Two to four weeks. The output is a list: keep, drop, replace with standard, rewrite.
- Count the integrations, including the ones other teams built against the AX database.
- Decide the history question with finance and audit before anyone quotes the data workstream.
- Decide how much process change you want and write it down, so the partner is pricing the same project you are buying.
- Ask for the quote by workstream in the shape of the table above, with assumptions listed, not as a single number with a duration.
- Add the missing lines from the section above, then the contingency.
Done this way, the figure is usually higher than the first quote and far lower than the final invoice would have been. The phase-by-phase timeline that goes with it is in how long an AX 2012 to Dynamics 365 upgrade takes, and what to do while you decide is in AX 2012 R3 end of life: what breaks next.
If you want the assessment done by people who have carried AX 2012 systems across, our AX 2012 to Dynamics 365 upgrade service starts there, and the contact page is the quickest route to a conversation.
Frequently asked
Is upgrading from AX 2012 cheaper than a new Dynamics 365 implementation?
Usually, because the processes, master data and people are already in place, but not by as much as the word “upgrade” suggests. If the AX system is heavily customised and the business wants to change its processes anyway, a reimplementation that migrates balances and open items can cost about the same and leave a cleaner system.
Can we reduce the cost by keeping all our customisations?
No, it usually raises it. Every customisation has to be rewritten as an extension and then tested on every Microsoft update. Dropping the ones the standard product now covers is the single biggest saving available on most projects.
Does the licence cost change when we move to Dynamics 365?
Yes. AX 2012 was a perpetual licence with annual enhancement; Dynamics 365 is a subscription per user per month, with different user types. Model it over five years alongside the hardware and support you stop paying for, and treat it separately from the project cost.
How much should we budget for after go-live?
Plan for hypercare through the first month-end and year-end, a regression pack for update waves, and a support arrangement (internal, partner, or both). On most projects that first year of ownership is a meaningful share of the project figure and is best agreed before go-live, not after.
