The short answer: choose by who owns the outcome after go-live. A fixed-scope SAP AI consultancy owns the deliverables in its statement of work. Staff augmentation adds people you direct, so you own the outcome. By contrast, an in-house hire makes the capability yours for good, but only once you have hired. An AI forward-deployed engineering team ships the system and owns how it behaves in production. So, before you compare rates or CVs, decide who will answer for the agent or model once real users depend on it. This guide compares the four options, says when each one wins and ends with eight questions to ask any SAP AI implementation partner.
First, one definition. An AI forward-deployed engineer (AI FDE) ships software on your systems and owns it in production, unlike an adviser who delivers a roadmap. Our AI forward-deployed engineers page covers the role in full.
The four options side by side
Read the table across a row to compare one question, then read the section below it that matches your situation.
| Question | AI FDE team | Fixed-scope SAP AI consultancy | Staff augmentation | In-house hire |
|---|---|---|---|---|
| What you receive | A production release, plus the code, runbook and evaluation suite | The deliverables in the statement of work, such as a roadmap or a scoped build | Individual contractors who join your team | Permanent people in your organisation |
| Who is accountable | The team, for the system in production | The provider, for the agreed deliverables | You; the contractor supplies capacity | You |
| SAP depth | Can include an SAP integration engineer | Varies by provider | Depends on the individual | What you hire for |
| AI engineering depth | AI engineers who build and evaluate the system | Varies by provider | Depends on the individual | What you hire for |
| Speed to first production release | 8–12 weeks from kickoff | Set by the statement of work | Depends on how fast you can direct the work | Starts after hiring and onboarding |
| Knowledge transfer | A trained client team at hand-over | Usually a handover document near the end | Depends on the individual | Stays with you by default |
| Risk | Release gates check quality and access before go-live | Running it after go-live can fall outside the scope | You carry evaluation, access and run decisions | The capability depends on a few people |
| Best for | A use case that must reach production and stay there | A clear, stable scope, or strategy before build | A known skills gap in a team that already ships AI | Long-term AI capability you want to own |
These are typical patterns, not rules, and individual providers vary. For the SAP build version of this comparison, see our SAP forward-deployed engineers page.
When each option wins
An AI FDE team
Choose an embedded team when a use case has to reach production in your SAP processes and nobody in-house can own evaluation, access and the run yet. For example, a pilot that has been almost ready for two quarters needs an owner more than it needs extra effort. Our page lists the situations where an AI FDE team is the right call.
A fixed-scope SAP AI consultancy
Choose a fixed scope when you can write the deliverable down today and it will not move. It also fits when you need an assessment or a roadmap before you commit to any build. However, check who will run the result after go-live, because a fixed scope may end at go-live.
Staff augmentation
Choose staff augmentation when you already have an AI team with its own evaluation practice and SAP access patterns, and you simply need more hands. As a result, your leads direct the work and carry the outcome, which is exactly what you want.
An in-house hire
Choose to hire when AI will be a lasting core capability and you can recruit and keep the people. Also consider a combination: an embedded team ships the first production release and trains your new hires to run it.
When Altivate is not the right fit
An embedded AI team, from us or from anyone else, is more than you need in three cases. First, you only want a strategy or a roadmap, with no build yet. Second, you only need extra capacity that your own leads will direct. Third, your team already ships AI on SAP with its own evaluation and run practice. In addition, if the open question is how the SAP process itself should work, start with our SAP forward-deployed consultants instead. For more on how the three roles fit together, see our forward-deployed teams page.
Why AI on SAP breaks the staff-augmentation model
Staff augmentation works well when the problem is capacity. AI on SAP also raises three problems that more people do not solve, because each one needs an owner.
- Evaluation. Someone has to decide when the agent is good enough. In practice, that means measuring quality against labelled real cases from your business before go-live, and keeping that set current.
- Data access. The agent has to reach SAP data, other systems and documents under the right identity. Therefore, someone has to choose between the user’s own SAP authorisations and a scoped technical user, and review that choice before release.
- Authorisations to act. At first, a person should approve every write or posting to SAP. Someone also has to watch usage, cost and errors, and be able to switch the agent off.
A contractor can do any of this work. However, a contractor you direct does not own it, so the decisions fall back to your team. That is why these are outcome problems, not capacity problems. For how we design these controls, see AI governance, evaluation and security.
Engagement shape at Altivate
In practice, a team is two or three engineers who work inside yours, for example an AI engineer and an SAP integration engineer. An engagement runs 8–12 weeks from kickoff to the first production release. Before go-live, the work passes four release gates: an authorisation review, an evaluation test set, human approval of writes, and monitoring with a way to switch the agent off.
At hand-over, you keep the code and a runbook in your own repository or SAP BTP subaccount. You also keep the evaluation suite and a team we have trained to run it. Then, if you want continuity, we can keep it running as managed AI operations under an AMS contract. For the full flow, see our AI FDE page.
Eight questions to ask any SAP AI implementation partner
Ask these before you sign, whichever option you choose. In short, if the answer to most of them is “you do”, you are buying capacity, not an outcome.
- Who owns the evaluation set, and does it use labelled real cases from our business?
- Which identity will the agent use in SAP: each user’s own authorisations or a scoped technical user?
- Who reviews that access before each release?
- Who approves writes and postings to SAP at first, and who decides when that changes?
- Who watches usage, cost and errors after go-live, and who can switch the agent off?
- What do we keep at hand-over, and where will the code, runbook and evaluation suite live?
- Which of our people will you train to run it?
- Finally, at exit, can we run it ourselves or move the run elsewhere without a rebuild?
For the wider picture of what SAP delivers, see SAP Business AI and AI for SAP. If you are also choosing a build route, read Joule Studio or a custom agent on SAP.
Frequently asked questions
What does an SAP AI consultant do?
A consultancy typically delivers advice and a roadmap, and sometimes a scoped build. By contrast, an AI FDE team builds the system and owns it in production.
Can we combine these options?
Yes. For example, you can hire in-house while an embedded team ships the first release and trains your new hires to run it.
Compare the options on your own use case
First, bring one use case and the SAP data it touches. We will walk through the four options with you, including whether an embedded team is the right fit.
