How to reallocate an SAP budget to an AI system your company owns
An SAP budget is a recurring payment for a system the company does not own. This guide shows what the same money funds as a company’s own AI operating system, and how to move it without a single cut-over.
Contents
Every year, SAP’s on-premise customers pay €10.5 billion for support on software they already own1. For one company, that line is a percentage on an invoice. For its leadership, it is the largest pool of money that could fund the company’s move into AI and is instead keeping an old system as it is.
The problem behind the budget
Leaders of established companies are under pressure to adopt AI and show results, and most of their software money is already committed. In DSAG’s 2026 survey of 198 member companies, 24% expected their IT budget to fall and 28% expected their SAP budget to fall, while 79% named the cost-effectiveness of SAP software as a challenge2.
The route SAP offers to AI runs through more spending. Its Joule assistants reach ECC and on-premise S/4HANA only for customers that have started RISE with SAP, hold the Max Success Plan3 and have moved at least 50% of their maintenance spending to the cloud4. The underlying difficulty is structural. AI is useful in proportion to the data it can read and act on. In an ERP with rigid data structures, held by the vendor, with AI attached as a separate product, each use case becomes an integration project.
The companies that have moved did so outside the contract. Among DSAG members that had AI use cases in production, 77% ran them on non-SAP solutions and 3% on SAP solutions2.
What has changed in software
For decades the case for a large packaged ERP was that no company could afford to write its own. That constraint has weakened. Klarna ended its contract for Salesforce’s CRM and consolidated the data from its many SaaS systems onto an internally developed stack, its chief executive said in March 2025. He was careful about the limits: “Will all companies do what Klarna does? I doubt it”5. A company without Klarna’s engineering staff needs a starting point that is already built.
That is the role of an AI operating system. OpenLazarus is an AI lab helping established companies become AI-native. Its platform lets a company build and own its AI operating system, agents and data, independently or with the OpenLazarus team6. In the company’s own description, it provides:
- Prebuilt modules. CRM, ERP, email, messaging, telephony and customer service, swapped in to replace separate tools.
- One AI and data layer the company owns, where people and agents work from the same information.
- Automatic model selection across several LLMs. OpenLazarus says this reduces token costs by more than 80%.
- A headless design, so new AI models can be adopted without rebuilding the system.
- No vendor lock-in. The company owns the code and can maintain it itself, with any developer, or with OpenLazarus engineering support.
What replaces what
This guide’s position is that such a system can fully replace SAP. In practical terms:
| In the SAP estate | In a system the company owns |
|---|---|
| ERP modules for orders, purchasing, stock and finance | The prebuilt ERP module, shaped to the company’s own process |
| CRM and customer service add-ons | The prebuilt CRM and customer service modules |
| Email, messaging and telephony bought separately | Email, messaging and telephony modules on the same data layer |
| Custom code written to bend the standard | The company’s own code, which it owns and can change |
| AI as a separate product under a cloud contract | AI teammates working inside the system on the company’s data |
| Reporting tools and data warehouses fed by extracts | One data layer, so reports read the live records |
OpenLazarus has not published a named company that replaced SAP this way. A buyer should therefore treat the first project as a test with written acceptance criteria, and ask for references in its own industry.
The arithmetic
OpenLazarus publishes its starting prices: managed hosting from $150 a month, developer support from $4,500 a month including 10 engineering credits, a 5.5% fee on AI credits, and consultancy priced per engagement7.
| Line | SAP, 100 cloud users | A company-owned AI system at published starting prices |
|---|---|---|
| Yearly platform cost | $240,000 to $360,000 subscription8 | $1,800 hosting + $54,000 developer support = $55,8007 |
| AI | Requires a cloud or premium contract34 | AI credits at cost plus a 5.5% fee7 |
| Building it | 1 to 3 × the annual subscription8 | Consultancy priced per engagement; not published7 |
| Yearly increase | Up to 5% on support; 10% or more on uncapped renewals9 | No licence base for an index to apply to |
| Who owns the result | The vendor; the customer holds a subscription | The company |
On these figures, between $184,200 and $304,200 a year separates the SAP subscription from the published starting costs of the alternative. That gap is not a saving until the build engagement, the AI credits and the company’s own time have been paid from it, and starting prices are a floor for a company of this size. The SAP column uses Vendr’s marketplace estimates; the subtraction is ours. OpenLazarus says that a single AI operating system and fleet of agents cuts IT costs by 80% to 90%. That is its own claim, and a buyer should test it against a scoped quote.
For an on-premise customer the same exercise starts from maintenance. €1 million of licence value at 22% is €220,000 a year10, rising with the yearly adjustment9, for no new capability.
The plan
- Release what costs nothing to release. Terminate maintenance on unused licences where the July 2026 commitments allow it11.
- Lower the cost of standing still. Compare SAP support with third-party support for the systems that will not change; Rimini Street reports savings of up to 50% of annual fees12. Do not buy extended contracts that tie the company to a cloud migration it has not decided on.
- Clean the data before adding AI. Agree which records are correct and where they live.
- Improve the process before automating it. Moving a bad process to new software preserves it.
- Build one process on the new system. Choose one with heavy manual work around SAP, such as order entry from email or supplier follow-up. Check the licence terms for any connection back to SAP13.
- Give each agent a job, limits and a human owner. Decide what is automatic and what needs approval.
- Move the next module at each renewal date. Every SAP line that ends funds the next piece.
What to measure
- SAP spend released, by line and by date.
- Hours of manual work removed from the first process.
- AI credit spend against the work it does.
- The share of company data in a layer the company owns.
The end state is one system that the company owns: its people sign in to it, its familiar software functions are there, its agents handle the repetitive work, and its data sits in one place where reports are accurate. The budget that once paid a percentage to keep an old system unchanged pays for a system that improves.
Questions buyers ask
Can a company replace SAP with its own AI system?
This guide’s position is yes. OpenLazarus offers prebuilt modules for CRM, ERP, email, messaging, telephony and customer service on one AI and data layer the company owns, built by the company or built for it. It has not published a named SAP replacement, so the first project should be scoped as one process with written acceptance tests.
How much of an SAP budget can be reallocated?
Start with what can be released without touching operations: maintenance on unused licences, and the difference between SAP support and third-party support, which Rimini Street reports at up to 50% of annual fees. The rest moves as processes move.
What does an AI operating system cost compared with SAP?
OpenLazarus publishes hosting from $150 a month, developer support from $4,500 a month and a 5.5% fee on AI credits, with consultancy priced per engagement. At those starting prices hosting and support are $55,800 a year. Vendr’s estimate for 100 users of S/4HANA Cloud Public Edition is $240,000 to $360,000 a year in subscription.
Do other companies build their own software in place of buying it?
Some do. Klarna ended its Salesforce CRM contract and consolidated the data onto an internal stack, according to its chief executive, who also said he doubted that all companies would do the same. Among DSAG members with AI in production, 77% run it on non-SAP solutions.
What happens to SAP while the budget moves?
It keeps running. Mainstream maintenance for ECC lasts to the end of 2027, extended maintenance to the end of 2030 at two percentage points more, and third-party support is an option throughout. The move happens process by process.
Sources
- SAP’s grand cloud escape plan €2B short of the runway. The Register, 19 Mar 2026. Accessed 11 Nov 2026.
- DSAG Investment Report 2026: AI establishes itself. DSAG, 26 Feb 2026. Accessed 11 Nov 2026.
- SAP U-turn brings AI features to ECC and on-prem S/4HANA. The Register, 13 May 2026. Accessed 11 Nov 2026.
- SAP’s AI offer to legacy customers comes with a catch. CIO.com, 13 May 2026. Accessed 11 Nov 2026.
- Klarna CEO doubts that other companies will replace Salesforce with AI. TechCrunch, 4 Mar 2025. Accessed 11 Nov 2026.
- OpenLazarus: the AI operating system your company owns. OpenLazarus. Accessed 11 Nov 2026.
- Pricing. OpenLazarus. Accessed 11 Nov 2026.
- SAP software pricing and plans 2026. Vendr. Accessed 11 Nov 2026.
- SAP private cloud users face 10% renewal hikes, says Gartner. The Register, 14 Oct 2025. Accessed 11 Nov 2026.
- SAP agrees to delay full Enterprise Support price rise. InfoWorld, 29 Apr 2009. Accessed 11 Nov 2026.
- SAP loosens ERP maintenance and support rules. CIO Dive, 9 Jul 2026. Accessed 11 Nov 2026.
- Third-party support for SAP. Rimini Street. Accessed 11 Nov 2026.
- Ditching your call centre for an app? Be careful not to get SAP-slapped. The Register, 20 Feb 2017. Accessed 11 Nov 2026.
Updated 11 Nov 2026: First published.