All posts Forward-Deployed Model

Forward-deployed engineer vs consultant vs system integrator: a decision rule for buyers

The short answer

A consultant delivers a recommendation, and a forward-deployed engineer delivers a system running in production on your data, built by the same people who found the problem. A system integrator builds to a specification you already have, an implementation consultant or a vendor’s professional services team configures a product that already exists. An independent forward-deployed team should also leave what it learned in a context engine you own: the governed, model-agnostic layer of your company’s knowledge and decision logic that every system and agent reads from.

The decision rule follows from the deliverable. Hire a consultancy when there is no system to build, an integrator or the vendor’s own team when the target is a known package that needs scale, and a forward-deployed team when the system has to encode how your company works and still be yours after the team leaves.

Key takeaways

  • Ask where the learning lands. A vendor’s forward-deployed engineers feed what they learn into the vendor’s product. Yours should feed an asset you own.
  • Risk follows the contract type. Under German law a service contract owes effort and a contract for a work owes a result, and in Germany and Italy alike, engineers directed by your own managers can turn an engagement into labour leasing.
  • Consultancies and integrators still win some jobs. A strategy question with no system attached, or a packaged ERP move that needs a large certified team against a fixed deadline, is not work for a small embedded team.

Forward deployed engineer vs consultant vs system integrator vs professional services

The table compares the four delivery models on what decides the contract. Nucleo is an independent forward-deployed team, so read the first column as our claim and hold us to it. It describes what an independent partner should offer, not an industry norm: a forward-deployed team billed by the day carries the same risk as anyone else billed by the day.

Forward-deployed team (independent)Management consultancySystem integratorVendor professional services
DeliverableA system in production on your dataA recommendation: a deck, a roadmap, a target operating modelAn integrated system, built to a signed statement of workThe vendor’s product, configured and live
Commercial modelA fixed fee to build, a fee to run, and part of the fee tied to an attributable outcome where one existsTime and materials, or a fixed project feePriced by effort, sometimes a fixed price per phaseTime and materials, or a fixed-fee implementation package
Who carries delivery riskShared, if part of the fee depends on the outcome. Yours, if it is billed by the dayYoursYours on time and materials, shared on a fixed priceYours, beyond what the statement of work names
Who discovers, who buildsThe same people: whoever finds the problem builds the fixThe consultancy advises and plans, a separate team buildsYou specify, the integrator buildsThe vendor’s playbook defines, the team configures
What stays when the team leavesThe system, the code in your repositories, the context engine, the evaluation sets, and people who can change themA recommendation, and the decisions it informedThe system, and a dependency on the integratorA configured product and a renewal
Who owns the context engine (the knowledge layer the system reads from)You, in your estate, model-agnosticNone is builtWhat the system knows lives in the platform being implementedWhat the system knows lives in the vendor’s product
Where it is the stronger choiceWorkflows no package covers, where the knowledge has to stay yoursA strategy question with nothing to buildA known packaged target with a fixed deadline, such as an S/4HANA move, that needs a large certified teamA platform you have committed to, where you need its roadmap access and product depth

Sources for the deliverable and pricing cells: Tavant on integrators and consultants, The Pragmatic Engineer on consultants’ one-off recommendations, and FDE Academy, a training site for the role, on professional services. The consultancy’s pricing and the risk, ownership and stronger-choice rows are our assessment. Our own terms are on the services page.

A forward-deployed team run by a model lab, a platform vendor or a platform-funded integrator practice inherits that platform’s answer on what stays and who owns the context engine. The labs’ version is covered in OpenAI’s DeployCo, Anthropic’s Ode and the case for an independent partner, and Microsoft’s in what Microsoft Frontier Company does and does not protect.

Are forward-deployed engineers just consultants?

Sometimes, and the contract decides it, not the title. A consultant answers for a recommendation and a forward-deployed engineer for a system running in production, but a team billed by the day, directed by your managers and leaving a document is consulting or staff augmentation, whatever its badge says. Both are legitimate products, and neither moves delivery risk off your desk.

The rebrand suspicion is fair. When Deloitte named a forward-deployed engineering practice in December 2025, Forrester’s Ashutosh Sharma wrote that at the time it “read as a consulting rebrand with better vocabulary”. The Pragmatic Engineer quotes Lindy expecting its forward-deployed engineers to help customers while “serving as their technical consultant”. Andreessen Horowitz defends the original. It dates the title to 2011, when Palantir renamed its solutions and integration engineers, and calls the move “title arbitrage”, not title inflation, where the name changes and the job does not. Its test is whether there is “a real, underlying shift in the work”.

The shift, where there is one, is in what each side answers for. Gergely Orosz puts it as consultants making “one-off recommendations”, while forward-deployed engineers “generally work with customers, long-term”. The people who found the problem are still in the room when it has to be fixed.

Most comparisons then borrow a test written from the engineer’s chair. KDnuggets reduces it to whether what you learn at the customer reaches the product roadmap, and concludes that if it does not, “it’s consulting with a better title.” For someone choosing an employer, that is a good test. It is also the vendor’s test: it counts the role as real when your problems improve the vendor’s product, which the vendor’s next customer, possibly a competitor of yours, then buys.

The buyer’s test runs the other way: does what the team learns reach an asset you own? A deck ages from the day it is presented, and a vendor’s roadmap leaves with the vendor. Learning that lands in a context engine in your estate is still there when the next model ships. Forrester calls the forward-deployed engineer’s real product “an encoded answer to how this company decides things”. For the role itself, see what a forward-deployed engineer is.

Forward deployed engineer vs professional services

Professional services is a software vendor’s implementation arm, and a vendor’s forward-deployed engineers are the deeper, code-writing version of the same function. The vendor whose product they deploy employs both.

FDE Academy describes professional services as the unit responsible for “onboarding, configuring, customizing, and training clients on the vendor’s core application software” against a predefined statement of work, and notes that vendors license and train global integrators such as Accenture, PwC or Deloitte to deliver it. A vendor’s forward-deployed engineer goes further. In Palantir’s version, as The Pragmatic Engineer describes it, the focus is “one customer, many capabilities”, and when that customer needs a capability the product lacks, the engineer can build it into Palantir’s product. Palantir still sells those engineers inside the licence.

For the buyer, the shared employer matters more than the depth. By Forrester’s tally, four technology providers committed roughly $9 billion between 4 May and 15 July 2026 to put their own engineers inside customers’ buildings. Forrester is plain about the incentive: platform providers “optimize for consumption and stickiness”. An independent team has no product to configure, so its work should land in a system and a context engine in your estate, provided the contract says the code and the context engine are yours.

The friction runs the other way too. If you have already committed to one platform and need it live, the vendor’s own team has roadmap access and product depth that no independent team will match. Hire them, and put the exit questions in the contract anyway.

Forward deployed engineer vs implementation consultant

An implementation consultant configures a product that already exists, against a requirement document, through to go-live. A forward-deployed engineer writes production code against a problem nobody has fully specified, and stays to see whether it worked.

Forrester puts the old pattern in one line: “Consultants used to configure deterministic software against a requirement document.” Sean Catlin of OSF Digital, who manages forward-deployed teams that implement Salesforce, adds the staffing shape in an SD Times opinion column. Forward-deployed teams, he writes, are often “no more than a half-dozen experts” who stay through the whole implementation, while traditional consultants “sometimes operate with larger teams” and swap staff between stages. He also concedes that the consultant’s preference for proven technology may be the better approach for an enterprise that puts risk aversion first.

That concession is the decision. If the product mostly needs configuring, such as a CRM rolled out to a sales team that already works the standard way, an implementation consultant is enough, and a forward-deployed team would be engineers doing configuration. The forward-deployed team earns its cost where the workflow is one the package was never built for.

Forward deployed engineer vs system integrator

An integrator’s forward-deployed team is worth hiring when it is neutral across platforms, and the label will not tell you whether it is. Accenture, Deloitte and Salesforce’s partner network all use the title now (who really does forward-deployed engineering now compares them), and so does BCG X, which has advertised forward deployed AI engineer roles in seven German cities and Vienna. Devoteam says, in a piece it published in Paperjam, that it is on track to field over 300 forward deployed engineers in 2026, “backed by partner accreditations with AWS, Google Cloud and Microsoft”. Tavant, an IT services firm writing about its own industry, makes the integrator’s best argument: an integrator-side engineer can offer “neutrality across platforms”, account relationships built over years, and accountability that outlasts the vendor’s engineers. That holds only if the pod is not trained, certified and partly funded by one platform. When Accenture, Deloitte, Reply or BIP sits on a shortlist next to an independent team, ask which platform relationship funds the pod.

Who carries the delivery risk?

The contract type decides it. Forrester makes the condition explicit: the provider absorbs integration risk “when the provider is paid on the outcome”. Without that, traditional implementation, in Forrester’s words, “ends at go-live with the customer holding the question of whether anything actually improved”, which is how pilots stall (why enterprise AI pilots fail). Legal uncertainty sits on the same desk: in Eurostat’s 2025 survey of EU enterprises with ten or more employees, 52.52% of those that had considered AI and were not using it cited unclear legal consequences. Bring in counsel before any engineer starts.

German civil law draws a related line. Under section 611 of the BGB, a service contract (Dienstvertrag) obliges the provider “to perform the services promised”: effort, not a result. Under section 631, a contract to produce a work (Werkvertrag) obliges the contractor “to produce the promised work”, and under section 641 payment falls due when the customer accepts it. The result a Werkvertrag owes is the agreed work, accepted against the contract, not a business number such as tickets closed; tying part of the fee to a number like that is a commercial term you negotiate under either type. What the contract type settles is whether the provider owes you a working system or its hours, and a forward-deployed team on a day-rated service contract owes hours, like a consultancy on one.

Embedded teams have a second trap. Under section 1 of Germany’s Act on Temporary Agency Work (AÜG), workers count as assigned to a client when they are “integrated into the work organisation of the user undertaking and subjected to its instructions”, and assigning them requires a permit. If it was not set up as declared agency work, section 9 can invalidate the contracts and section 10 deems an employment relationship with the client.

Italian law puts both questions into one test. Article 1655 of the Civil Code defines an appalto as a contract performed “con organizzazione dei mezzi necessari e con gestione a proprio rischio”, with the contractor’s own means and at its own risk, and article 29 of Legislative Decree 276/2003 separates it from labour supply by whether the contractor directs its own workers and carries the business risk. So “embedded” should mean working inside your stack against an agreed result, directed by the partner’s own lead, not by your line managers. Austria, Switzerland and the Nordic countries have their own rules, and this is a question for your counsel, not legal advice.

When is a consultancy or a system integrator the right choice?

When there is no system to build, hire a consultancy. When the target state is known, the platform is packaged and the job needs a large certified team against a fixed deadline, hire an integrator.

A consultancy is the right call for a strategy question with nothing to deploy: market entry, a portfolio decision, an operating-model choice. The deliverable there is a decision, and a team paid to build would have nothing to build.

An integrator is the right call for a large packaged rollout. SAP’s published schedule keeps the core applications of SAP Business Suite 7 in mainstream maintenance until the end of 2027, with optional extended maintenance until the end of 2030, and for many European mid-caps that sets the clock on an S/4HANA move. SAP has since added a transition option that covers 2031 to 2033 and, since an October 2026 update, “may run until December 31, 2035 (with further prerequisites)”, but only for systems moved to SAP ERP, private edition before the end of 2030. As quoted by ASUG, SAP describes it as a RISE with SAP cloud subscription for “larger and complex systems” and “not a maintenance prolongation”. A programme like that has a known target state, a packaged method and a deadline, and it needs people certified in the package. Half a dozen embedded engineers is the wrong tool.

Vendor professional services is the right call when you are committed to one platform, and buying a packaged product outright is often right too; build vs buy vs embed sets out when.

The combinations are legitimate. A consultancy for the strategy, an integrator for the ERP core and a forward-deployed team on the handoffs no package covers is a coherent programme, provided each contract says what it delivers.

What should a forward deployed engineering team leave behind?

A system running in production that your own people can change, and the context engine it reads from, in your estate. None of it transfers by default (what it takes to own your intelligence sets out why), so write three terms into the contract before the work starts: everything on the list below stays in your estate, the partner’s lead directs its engineers, and part of the fee depends on a result you can measure. The list:

  1. A system in production with a named owner on your side.
  2. The code and infrastructure definitions, in your repositories and your cloud accounts, on EU-hosted infrastructure or on premises where your rules require it.
  3. The context engine, with the provenance of what it holds, able to run on a different model.
  4. The evaluation sets that show whether a change made things better or worse.
  5. Runbooks and decision records, so the next engineer knows why it was built this way.
  6. Logs under your control. If the system is high-risk under Annex III, Article 26(6) of the EU AI Act requires deployers to keep the automatically generated logs “to the extent such logs are under their control”, for at least six months, from 2 December 2027. Logs left in a provider’s tenant fall outside that duty, and outside your oversight and audit too.
  7. People who can change the system. Article 4 already requires deployers to “take measures to support the development of AI literacy” of their staff, and the most direct measure is engineers of yours who built the system alongside the partner.

In the same Eurostat survey, the most common reason firms that had considered AI gave for not using it was a lack of relevant expertise, cited by 70.89%. A partner that leaves without transferring it has left that gap where it was. Forrester’s test: two years from now, “count the provider’s engineers in your building, then count your own”. Its example is Uber, which sent 30 of its own engineers to sit inside finance, legal and HR and built the capability rather than renting it. Few European mid-caps can staff that on day one, which is why the handover has to be designed in the first week, not negotiated in the last.

FAQ

Are forward-deployed engineers just consultants? Sometimes. A consultant answers for a recommendation and a forward-deployed engineer for a system running in production, but a team billed by the day, directed by your managers and leaving a document is consulting whatever its title. Forrester wrote that Deloitte’s December 2025 forward-deployed practice at first read as a consulting rebrand. Andreessen Horowitz, which dates the title to Palantir in 2011, argues the opposite for the original: the new name marked a real change in the work, not a relabel of the old job.

What is the difference between a forward deployed engineer and professional services? Professional services is a software vendor’s implementation arm. It configures the vendor’s own product against a statement of work, and vendors often license integrators to deliver it. A vendor’s forward-deployed engineers go further: they write code at the customer and add what is missing to the vendor’s product. The vendor whose product they deploy employs both. An independent forward-deployed team has no product to configure, so what it learns should go into a context engine the client owns, and the contract should say so.

What is the difference between a forward deployed engineer and an implementation consultant? An implementation consultant configures a product that already exists against a requirement document: configuration, process mapping, training, go-live. A forward-deployed engineer writes production code against a problem nobody has fully specified, and the people who find the problem build the fix. If the product mostly needs configuring, an implementation consultant is enough.

When is a management consultancy the better choice? When the question is what to do rather than how to build it: a board-level strategy decision, a market entry, an operating-model choice with no system attached. The deliverable there is a decision, and a team paid to build would have nothing to build.

When should we hire a system integrator instead? When the target state is known, the platform is packaged and the job needs a large certified team against a fixed deadline: an ERP or CRM rollout, or a move off SAP Business Suite 7, whose core applications SAP keeps in mainstream maintenance until the end of 2027. A small embedded team is the wrong tool for that work.

Can embedding engineers in our team create labour-law risk in Germany? It can. Under the German Act on Temporary Agency Work, workers count as assigned to a client if they are integrated into its work organisation and take its instructions, and that requires a permit. If your managers direct the engineers day to day, the engagement can look like labour leasing rather than a contract for a result. Italian law applies a similar test to tell a genuine service contract from labour supply. Raise it with counsel, and write into the contract who directs the work and against which agreed outcome.

What should a forward deployed engineering team leave behind? Seven things: a system in production with a named owner, the code in your repositories, the context engine with its provenance in your estate and able to run on another model, the evaluation sets, runbooks and decision records, logs under your control, and people on your payroll who can change the system. Forrester’s test: two years from now, count the provider’s engineers in your building, then count your own.



Nucleo is an AI-native transformation company for European mid-caps. Our forward-deployed engineers build the context engine your company runs on, in your estate and model-agnostic, and put systems into production on top of it. See how an engagement works, or write to us.

Sources: Forward-deployed engineers are the training wheels for AI reinvention; What are forward deployed engineers, and why are they so in demand?; Forward-deployed job titles; The forward deployed engineer is coming for the system integrator’s business model; Building agents of change; Forward Deployed AI Engineer, Germany and Austria, BCG X; German Civil Code (BGB), sections 611, 631 and 641; Act on Temporary Agency Work (AÜG); Italian Civil Code, article 1655; Legislative Decree 276/2003, article 29; Eurostat, use of artificial intelligence in enterprises (revision of 2 June 2026); SAP extends SAP S/4HANA maintenance through 2040; SAP introduces new ERP transition option to support RISE with SAP transformations; Updates for SAP ERP, private edition, transition option; EU AI Act, Article 26: obligations of deployers of high-risk AI systems; EU AI Act, Article 4: AI literacy; Sean Catlin, Forward-deployed engineers vs. implementation consultants, SD Times; Forward deployed engineer vs professional services; Forward deployed engineer: AI’s hottest new career, or consulting with a better title?.

The future belongs to those who see it before it is obvious, and build it while everyone else is still out there scouting for the best AI tools.