Every GSI has the same code assistants. What sets a bid apart now?

Moti Rafalin

November 16, 2026

Every firm bidding on modernizing your clients’ largest applications now has the same AI coding tools. What sets one offer apart is the platform those tools work from and the engineers who work alongside it, and that delivery model is taking shape in this modernization cycle.

Two questions decide who leads: what differentiates your modernization offer when everyone uses the same AI coding assistants, and what is the margin you can get on each forward-deployed engineer.

The GSI model is being rewritten

The cost of waiting to modernize is going up fast, and the cost of modernizing has never been lower. AI coding tools made that true by writing code quickly and cheaply. They also reached every competitor at the same moment.

That makes the coding assistant a baseline for every bid. The offers that stand out will be platform-led and delivered by forward-deployed engineers (FDEs): engineering teams working with a platform that gives AI coding tools the context to efficiently modernize large existing applications.

What LLMs lack

Past roughly 500,000 lines of code or 1,000 classes, modernization stops being a code problem and becomes an architecture problem.

When writing the code is trivial, the challenge becomes deciding what are the right domain boundaries given the business function being delivered, which dependencies are irrelevant and can be removed, and what needs to keep working as-is. Code assistants build new applications well — they are essentially optimized for greenfield applications.

However, on large existing applications, three fundamental gaps emerge:

  • Source code does not show what matters. It lists every dependency, but not which ones are critical, what actually runs in production, or where the real business domains sit.
  • Context windows run out. On applications of tens of millions of lines of code, working through the whole system is either impossible or prohibitively expensive.
  • The result has to be proven. Someone must show that the refactored or rewritten application behaves like the original and validate the original behavior was kept intact.

Without that foundation, an LLM rewrite can deliver a “new” application in a modern language that carries the same architectural technical debt. That outcome will cost you more and fail to deliver what the client is looking for.

The differentiation

vFunction, which powers AI to refactor and rewrite, gives AI coding tools the context, specifications and tests they need to modernize large existing applications. It does so by providing AI coding tools four inputs from the running application and its own deterministic proprietary data science:

  1. A system model of how the application actually runs in production.
  2. A target architecture the client defines, from a domain structure vFunction proposes from runtime behavior.
  3. A specification for each domain, backed by observed behavior, without the debt the client chose to remove.
  4. Integration tests from production behavior, to prove each change.

With these in hand, your developers’ AI coding tools work in a token-efficient manner, with architectural certainty and guardrails. The modernization becomes a series of specified, tested changes in the order the client’s team sets, rather than one large rewrite.

FDE economics: the $2M question

Revenue per engineer is where a delivery model shows up in the P&L. The combination of vFunction, an AI coding tool, and a forward-deployed engineer supports about $2M of revenue per FDE on modernization work. I encourage any GSI to test that model against its own numbers.

Three things move the arithmetic:

  • Each change is specified and tested against how the application runs, so a project needs fewer people on the architectural core.
  • Available engineers can staff the work. Refactoring has succeeded with teams new to the target application, without years of tribal knowledge.
  • The method repeats. There are examples of institutions that have run it across dozens of applications.

The AI-era delivery model is being set now. The GSI that defines it with the right platform leads the category, and that window won’t stay open.

The delivery stays yours

vFunction handles the architectural core. Deployment, integration, UI modernization, and everything around the refactoring is your work, across a wider set of engagements. The architectural core is where the combination of vFunction and AI coding tools compress the work. The rest is where your teams deliver, and it grows with every application that becomes viable to take on.

vFunction’s forward-deployed engineers work your engagements alongside you, and we are measured on whether the project lands. As a side note, for an application on AWS or moving to AWS, AWS funds the vFunction work through its ISV tooling program, so you can analyze the application first and then decide how to bid.

Where we fit

vFunction is a modernization platform for large existing applications. It gives AI coding tools the context and guardrails they need to refactor, rewrite, and modernize existing systems with speed, accuracy, and control. vFunction modernizes applications in Java, .NET, Python, C/C++, JavaScript, and PHP, but not mainframe or COBOL.

Keep learning: How a vFunction engagement works · The cost of waiting is going up

Tell us about the application. Bring us one real candidate from a client’s estate. We analyze it with you, you bid with certainty, and you deliver with us alongside you. Let’s Talk.

Moti Rafalin

CEO, co-founder

Moti Rafalin co-founded vFunction and serves as its CEO. He brings over 20 years of experience in the enterprise software market to his role, with a focus on infrastructure, applications, and security. Prior to co-founding vFunction, Moti co-founded and led WatchDox from inception until its acquisition by BlackBerry, growing the company over 20 consecutive quarters and establishing it as a leader in the secure enterprise mobility and collaboration space. Subsequently, he served as Senior Vice President at BlackBerry LTD. Before WatchDox he was a general manager of the application management business at EMC. He holds an engineering degree from Technion and an MBA from the Harvard Business School.

Let's talk about your application.

Accelerate engineering velocity, boost application scalability, and fuel innovation with architectural modernization. See how.