The latest phase of AI adoption is creating high demand for engineers who can do more than create basic prompts or generate code. Companies need experienced people who can enter an organization, identify problems that AI can solve, connect new technology to existing systems and turn ambitious demos into working business applications. These specialists, known as forward-deployed engineers (FDEs), are moving from a relatively obscure corner of enterprise software into the center of the AI industry’s commercial strategy.

Recent acquisitions of AI consulting and engineering firms by OpenAI and Anthropic are bringing this demand into sharper focus, as leading AI companies invest in organizations with the expertise to deploy their technology directly inside customer businesses. OpenAI’s agreement to acquire Tomoro and the acquisition of Fractional AI by an Anthropic backed enterprise services venture illustrate how valuable implementation expertise has become to companies seeking to move beyond selling AI models and into delivering working enterprise applications.

The stakes extend well beyond engineering recruitment or AI implementations. AI vendors can build increasingly sophisticated models, but customers still face fragmented data, undocumented processes often called “tribal knowledge”, regulatory restrictions and uncertainty about which applications justify the investment. Forward-deployed engineering offers a way to close the gap between the capabilities of AI and the operational realities of the organizations expected to use it.

For technology providers, the approach may shorten sales cycles, increase adoption and create more durable customer relationships. For enterprises, it promises a more direct route from experimentation to measurable returns.

Yet the growing enthusiasm raises a consequential question. Should AI companies embed their own engineers with customers, accepting the cost and complexity of becoming part software vendor and part consultancy? Or should they retain a more conventional professional services model, using targeted discovery, workshops and rapid prototyping to identify problems before committing engineering resources?

Conversations I had this past week with two technology companies in Kaunas, Lithuania’s second largest city and a rapidly growing global AI hub, reveal why the model is attracting attention, and why even in the advanced edges of the AI market there is disagreement about how far it should go and what happens after the AI demonstration ends.

Why Frontier AI Companies Are Buying Engineering Expertise

In May 2026, OpenAI announced the creation of the OpenAI Deployment Company, a new business backed by more than $4 billion in initial investment and supported by 19 investment firms, consultancies and systems integrators. OpenAI simultaneously announced an agreement to acquire Tomoro, an applied AI consultancy whose approximately 150 engineers and deployment specialists would provide the venture with an established delivery organization from its inception.

Continuing this momentum, in July, the OpenAI Deployment Company announced an agreement to acquire Northslope, another applied AI firm with experience implementing AI in complex enterprise environments. The company’s stated objective is to bring engineers directly into customer organizations, connecting AI models to proprietary information, existing applications, security controls and operational processes.

Not to be outdone, Anthropic is pursuing a related approach through Ode with Anthropic, an enterprise AI services business established with Blackstone, Hellman & Friedman and other investors. Introduced under the Ode name in July 2026, the company was built around Fractional AI, an applied AI services firm acquired in May. Ode subsequently acquired Casper Studios in August, adding expertise in deploying Claude inside the applications and workflows that companies already use.

The commercial logic for these tech firms is compelling. A model provider earns revenue when its technology is used, but an enterprise that cannot integrate AI into its business processes has little reason to expand its consumption. Engineers who identify applications, complete integrations and establish working systems can turn a limited experiment into sustained demand. Model providers are in this way moving closer to their customers’ operating environments, acquiring both technical expertise and firsthand knowledge of the business problems their technology must solve.

Startups Are Adding FDEs To Their Customer Acquisition Models

For Kaunas-based startup Ace Waves, the forward deployed engineer is not a supplementary resource brought in when a customer encounters implementation difficulties. The engineer is part of the company’s commercial proposition from the beginning.

The startup develops AI agents that handle customer service interactions for large consumer businesses. Its technology uses agents that go beyond answering basic questions, and can process subscription cancellations, modify customer accounts, retrieve information from payment systems and execute transactions through existing support channels.

Delivering that capability requires considerable customization. A subscription business, for example, might permit an AI agent to offer a discounted plan before canceling a customer’s account. Another might require immediate cancellation when the risk of a payment dispute exceeds a predetermined threshold. The underlying AI technology may be similar, but the business rules, integrations and operational safeguards can differ considerably.

Ace Waves founder Antanas Bakšys explained that approximately one third of its workforce is engaged in forward deployed engineering. The company typically assigns a dedicated engineer, supported by other team members, to an implementation. Its goal is to launch an initial working system during the first month, followed by refinement and expansion. The engineers participate before a contract is signed.

“Our FDEs are part of a sales team,” Bakšys explained. “We use this approach to demonstrate what our technology can accomplish inside a prospective customer’s actual operating environment, rather than asking the customer to purchase a platform based principally on demonstrations or promised capabilities.”

The company’s stated ambition is to automate 80% of a customer’s support workload and transform customer service from an expense into a source of revenue. A working pilot provides an opportunity to test the claim against the customer’s own systems and procedures.

The commercial model extends into pricing. Ace Waves says its contracts are based on successful customer service resolutions, aligning its revenue with the performance of the AI agents it deploys. Peformance-based pricing, rather than an hourly or project-based pricing approach, is often a counterpart to the FDE model. This gives the company a financial incentive to understand customer operations before implementation begins, rather than be motivated to bill more hours for complicated implementations or ration precious engineer talent in a fixed-price approach.

For Consulting and Solution Firms, The FDE Model Needs Reinvention

Not all AI implementation companies are of the same mind when it comes to using FDEs as part of AI projects. Gritmind approaches the same problem from a different direction.

The Kaunas based software engineering company works with enterprise customers on application development, modernization and increasingly, AI related projects. Its engineering operations are based in Lithuania, with its sales, business development and commercial expertise in the United States.

Gritmind’s representatives recognize the demand for forward deployed engineering. They are less convinced that establishing a dedicated deployment organization or assigning individual engineers to customers represents a fundamental improvement over established software consulting practices.

“We don’t want to give our key engineers to someone,” company partner and Head of Engineering Ramūnas Zavistanavičius explained, referring to the prospect of assigning an individual engineer to a customer without a broader delivery structure.

Instead, Gritmind favors temporarily deploying small teams to investigate business problems, conduct workshops, establish technical requirements and determine which applications merit further investment. The company explains that this is part of its traditional consultative selling approach.

“We are not jumping to suggesting AI this, AI that,” Zavistanavičius said during the interview, explaining that the company first seeks to understand the customer’s operational difficulties.

Its approach begins with the business problem rather than a predefined AI solution. The process is familiar to experienced consulting firms. Gritmind has conducted discovery workshops and developed customer applications for years.

But AI is changing things even for traditional solution provider consultancies. What has changed is the speed at which those workshops can produce convincing technical demonstrations.

Previously, early customer discussions might have resulted in design documents, workflow diagrams or clickable prototypes. Generative AI and so-called “vibe coding” tools now permit Gritmind’s engineers to build functional prototypes with usable interfaces and real outputs within substantially shorter development cycles.

In one enterprise invoice validation engagement described during the interview, Gritmind built a small proof of concept using a real invoice and a contract. The demonstration helped communicate the proposed solution before the customer committed to a larger implementation.

The company’s representatives said they won the engagement after competing against three other vendors, attributing the outcome in part to the functional prototype they produced. The implication is significant. Customers increasingly have the opportunity to experience a proposed application before purchasing the full development project. In some cases, the customers themselves are coming to the technology firms with their own functional prototypes.

Gritmind argues that the underlying consulting discipline remains relevant. Engineers must still understand the customer’s business, determine what should be built and identify the appropriate technology. AI makes that process faster, but the company argues that It does not automatically make a dedicated forward deployed engineering organization necessary.

The Real Obstacle To AI Adoption Is Often Hidden Inside The Business

Despite their disagreement on the right mode to engage the customer prior to the start of an engagement, the most revealing connection is their independent identification of the same obstacle to enterprise AI adoption.

Both encounter organizations whose business processes contain undocumented knowledge that cannot be translated directly into software. Ace Waves describes this condition as customer service debt, and Gritmind refers to this as “tribal knowledge .”

Policies may be distributed through internal documents, individual employees may handle exceptions differently, and important business rules may exist only in the memories of experienced customer service representatives. An AI agent cannot reliably execute a procedure that the organization itself has never defined with sufficient precision.

The problem becomes more pronounced when the agent must take actions rather than provide information. Incorrectly answering a question creates one category of risk, but canceling the wrong subscription, issuing an unauthorized refund or modifying a customer account creates another.

Ace Waves’ FDEs work with customers to uncover these operational inconsistencies, document decision rules and connect the relevant systems before expanding automation before the contract is signed or in the very early stages to determine the value that the organization can bring.

Gritmind encountered a comparable challenge in its invoice validation engagement. The company described a process where experienced employees rely on knowledge of vendor contracts, historical transactions and internal practices when deciding whether invoices are valid.

Existing documentation does not always capture the nuances required to make those decisions. Gritmind’s team has been exploring how historical records, contract information and documented procedures can help establish rules for automated invoice validation.

As part of this engagement process, an engineer who understands the customer’s business may discover that the appropriate solution combines AI with conventional software, human approvals and deterministic rules. From this perspective, the solution succeeds when the business problem is solved, not simply when the customer consumes more AI services.

Why The Demand For Forward Deployed Engineers Is Growing

Forward deployed engineering did not originate with generative AI. Palantir popularized the model through engineers who worked closely with customers to adapt its software to complex operational requirements. The rise of AI has brought a fresh commercial urgency to the approach.

General purpose AI is different than typical enterprise software development, web or mobile app creation or business intelligence dashboards.

The same AI frontier models and technology might support contract analysis, customer service, software development, procurement, financial reconciliation or internal research. Determining which application warrants investment requires knowledge of the customer’s business and the technical possibilities of the underlying model.

That creates a difficult sales problem. Customers may recognize AI’s broad capabilities without knowing which use cases deserve priority or how to translate them into functioning applications.

Forward deployed engineers can bridge that gap through a combination of technical development and business discovery. They can identify opportunities that customers have not considered, test whether those opportunities are technically feasible and establish whether the resulting applications deliver a measurable financial return.

Successfully building a high-return business case not only can win more deals, but also put more weight behind the arguments that AI is providing real business value.

In February 2026, OpenAI announced its Frontier Alliances with Accenture, Boston Consulting Group, Capgemini and McKinsey. The initiative combines the company’s engineering expertise with the consulting firms’ experience in organizational change, systems integration and enterprise transformation.

Forrester subsequently documented major investments in forward deployed engineering by Amazon Web Services and Microsoft, reflecting a broader industry commitment to deployment capabilities rather than model development alone.

The distinction matters commercially. A company may purchase AI subscriptions for thousands of employees yet realize limited operational value if the technology remains disconnected from its most significant business processes. In other words, companies are attempting to reward the use of AI by employees to improve outcomes, but employees lack sufficient knowledge and experience to implement AI projects that have significant, high returns.

An AI vendor with experienced FDEs that help customers redesign those processes has a more direct opportunity to demonstrate value, secure recurring revenue and develop stronger relationships. For enterprises, the attraction lies in reducing the distance between identifying an opportunity and delivering a working application.

Can Forward Deployed Engineering Scale?

The two companies’ approaches reveal a tension that may determine how the emerging AI services market develops.

Ace Waves has deliberately combined proprietary software with a substantial engineering services organization. Its engineers use the company’s own platform to build and orchestrate customer service agents, creating opportunities to reuse technical components and operational knowledge between engagements. From this perspective, its FDE approach is more like a form of business development. This also helps distinguish its business from a conventional consultancy that develops each customer application independently.

The model still requires substantial human involvement. New customers introduce unfamiliar systems, undocumented procedures and operational exceptions that engineers must investigate. As the customer base expands, maintaining sufficient engineering capacity could become a significant operational and financial challenge.

Gritmind faces a related question from the opposite direction. AI tools allow its teams to develop applications faster, potentially reducing project durations and changing the economics of traditional hourly engineering services. But deploying billable engineers to customers without any promise of future business might be too risky, especially if the implementation can be done independently or with other providers.

The next phase of enterprise AI adoption may be shaped less by another dramatic advance in model capabilities or some new agentic tool than by the human-to-human engineering work required to make existing capabilities useful. As technology firms explore how best to provide this human touch, whether the business can organize its people, processes and systems so that the technology can perform that task reliably, economically and at scale.