New framework enables enterprise organizations to evaluate whether AI has transformed actual work delivery, rather than just measuring tool usage, proven through extensive internal deployment.
AUSTIN, Texas — September 29, 2026 — Growth Acceleration Partners (GAP) has released its AI Impact Framework, a new approach to measuring how work gets delivered and whether AI is materially changing business outcomes. Available immediately to clients, the AI Impact Framework evaluates actual results.
The proprietary framework addresses a growing challenge for enterprises investing in AI: most companies can measure whether employees are using AI, but cannot tell whether AI has transformed the work or merely been layered on top of it. These organizations cannot distinguish an engineer whose output has fundamentally improved from one whose workflow looks exactly as it always did, with AI simply running alongside it.
GAP’s CEO and Cofounder Joyce Durst said the AI Impact Framework exists because GAP sells outcomes, not tools. And as an AI transformation company, GAP needed a way to prove AI was changing the impact of those outcomes.
“AI adoption is easy to count, but AI impact is harder to prove,” Durst said. “License utilization, prompt volume and self-reported adoption measure activity. They do not show whether work is being delivered faster, automation is being created or capabilities are becoming reusable. If you can’t measure that, you can’t manage it.”
The AI Impact Framework is available now to GAP clients through workforce assessments, benchmarking methodologies and development programs designed to move teams from AI-assisted work toward increasingly autonomous delivery.
How the AI Impact Framework Works
Under the framework, each participant submits evidence of AI use from the prior six months. This evidence must be real artifacts, not hypotheticals. This includes production code shipped with AI, custom agents in daily use, automated test suites and delivery pipelines, large-scale code migrations, internal tools other people now rely on, or reusable systems that reduce work for an entire team. What an employee believes they could do with AI does not count; only what they have actually built, automated or delivered does.
The framework uses a multi-agent workflow to evaluate what employees have actually built, automated, delivered, or made reusable, rather than relying on self-reported AI capabilities. The evaluation is conducted across multiple criteria, 22 in the case of engineering.
For each evaluation criterion, three AI agents work together as a unified pipeline:
- An evaluator agent assesses the evidence and assigns an initial score.
- An adversarial agent challenges unsupported claims, inflated attribution, and reasoning that is not supported by the evidence.
- A referee agent considers both assessments and determines the resulting score for that criterion.
This three-agent process runs independently across all evaluation criteria. Once all criteria have been evaluated, a fourth agent synthesizes the cumulative results and determines the employee’s final level. Human-in-the-loop oversight is incorporated into the overall process.
The AI Impact Framework is intentionally designed to challenge each answer of the assessment. The evaluator agent scores the evidence, the adversarial agent tests it, and the referee resolves the disagreement before human validation. Those scores aggregate into an AI maturity classification. Managers and project leaders then validate the classification, with disputed ratings escalated for additional calibration. Each stage of the process is documented to create a traceable assessment.
“The adversarial layer is designed to address a common weakness in AI maturity assessments: relying heavily on self-reported adoption, rather than demonstrated evidence of impact,” said GAP’s Chief People Officer Andrea Mena. “Our framework intentionally challenges the evidence before a score is finalized. GAP designed the framework to produce a more defensible measure of AI capability, rather than a flattering measure of AI activity.”
After the evaluation, participants are classified across four maturity levels, ranging from Level 0, where there is no meaningful AI integration, to Level 3, where employees demonstrate the ability to create and operate AI agents, automated workflows and reusable systems that generate impact beyond their individual work.
GAP is also using the AI Impact Framework to identify what capabilities individuals need to develop next. A personalized roadmap is included in every employee’s assessment to give specific direction and support for how to use AI to make more powerful, impactful outcomes for the company. Rather than treating AI transformation as a technology deployment, GAP uses the framework and roadmap to establish a baseline, identify capability gaps, and create development paths that move employees from basic AI assistance toward increasingly autonomous ways of working.
Proven Internal Deployment & Case Study Results
Before launching the AI Impact Framework to commercial clients, GAP first gave the assessment to its own workforce of approximately 500 employees. The initial assessment in early 2026 found roughly 11% GAPsters (as employees are affectionately known) had reached the highest maturity level, defined not by how often they use AI, but by what they build with it: AI agents, automated workflows and reusable systems that other teams can adopt, that keep producing value long after a single project ends, and that change how work gets delivered across the organization rather than for one person or one task.
Company-wide assessments revealed a critical divide between routine AI usage and true operational transformation. GAP launched extensive, role-specific training across all departments to close the AI gap within its own workforce, equipping employees to turn daily usage into measurable impact. Driven by hands-on skill development built to move staff from basic adoption to advanced systems creation, GAP achieved a dramatic shift by July 2026:
- Approximately 41% of GAPsters reached Level 3, demonstrating the ability to build agents, automated workflows and reusable AI systems with impact extending beyond an individual employee or project.
- Approximately 48% of GAPsters reached Level 2, using AI meaningfully but primarily generating individual or project-level gains, rather than changing how work is delivered at scale.
“The number that should concern leaders isn’t how many people aren’t using AI. It’s how many are using it every day without fundamentally changing what the customer receives,” Mena said. “That’s the gap most enterprise AI dashboards can’t see. We built GAP’s AI Impact Framework because we wanted to measure outcomes, not activity. But we also needed to test it on ourselves before asking clients to do the same.”
Every GAPster’s evaluation roadmap feeds directly into a personalized development tool. For each assessment cycle, this educational machine takes GAPster’s own evaluation feedback — the specific language a reviewer or the AI evaluator used to explain what’s missing — and turns it into a concrete, cycle-specific action plan matched to internal GAP Academy courses and/or available external courses we can have access to. Feedback becomes an actual plan, automatically, every cycle.
GAP Defines the Autonomous Engineer
For software engineering roles, GAP refers to the highest level of AI-enabled capability (Level 3) as the Autonomous Engineer. An Autonomous Engineer is a senior engineer who has moved from doing the work by hand to directing a system that does much of it: orchestrating AI agents to code, test and transform software at scale, while personally owning the decisions AI cannot be trusted to make: architecture, planning, quality governance and business outcomes. The engineer’s value shifts from writing every line to designing the system that writes it, and from executing tasks to guaranteeing results.
The distinction is intentional. Autonomy applies to the engineering system, not to the absence of human accountability.
Technical proficiency alone, however, is not sufficient to reach the highest maturity level. Under GAP’s Human + AI Autonomous Engineering model, employees must also demonstrate the judgment and human capabilities required to govern increasingly autonomous systems. Deficiencies in communication, collaboration, resilience, judgment or accountability can prevent an otherwise technically qualified employee from reaching Level 3.
“An engineer who can orchestrate sophisticated AI agents but cannot set appropriate guardrails, communicate trade-offs, collaborate effectively or take ownership of outcomes does not meet GAP’s definition of an Autonomous Engineer,” Durst said. “The more capable the tools become, the more everything depends on the person deciding how to use them. That is not a skill AI can replace. The human in the loop is still what matters most.”
The model is rooted in GAP’s Human + AI Autonomous Engineering approach, where engineers combine deep technical expertise with advanced AI capabilities. AI agents execute and accelerate portions of delivery, while engineers retain responsibility for architecture, planning, guardrails, judgment, quality and business outcomes.
The distinction from conventional AI-assisted development is the division of labor. AI is not a helper offering suggestions a person accepts or ignores; it executes meaningful portions of delivery under a human’s direction and guardrails. The engineer’s judgment — about what to build, what “good” looks like, where the risks are, and when to overrule the system — becomes the scarce, decisive input. As agents take on more execution, the human’s role concentrates rather than disappears.
“Our goal is to enable people and AI together to deliver results neither could achieve alone,” Durst said. “AI brings speed and scale. People bring judgment, context and ownership. The engineers who win the next decade won’t be the ones who resist AI or the ones who hand everything to it; they’ll be the ones who know exactly where each belongs.”
Linking AI Maturity to Workforce Advancement
GAP has stopped treating AI capability as a bonus skill and started treating it the way it treats technical proficiency, leadership and business impact. Value derived from AI is something GAP expects, measures and rewards.
For GAP’s 2026 performance cycle, achieving Level 3 is required for both promotions and eligibility for merit-based salary increases. GAP is also incorporating AI capability into job-level definitions across engineering, finance, HR, recruiting, marketing and operations.
This requirement reflects a simple principle: GAP won’t ask clients to transform how their people work with AI if it treats that same capability as optional for its own workforce. GAP’s leaders believe AI capability will become as fundamental to career progression as technical proficiency, leadership and business impact. And as AI systems become more autonomous, human judgment becomes more important, not less.
For clients, GAP’s AI Impact Framework is designed to show whether AI is producing isolated individual efficiencies or changing how teams operate, in addition to what they can automate and what the organization can ultimately deliver to customers.
For more information about GAP’s AI Impact Framework, visit www.WeAreGAP.com.