Jellyfish has introduced a suite of software engineering intelligence features designed to help organizations measure what AI coding tools and agents are actually changing across development teams. The platform combines AI usage, engineering activity, cost and delivery data to move beyond adoption metrics toward productivity, capacity and return-on-investment analysis.
AI coding assistants have moved quickly from experimental tools to a regular part of software development, but measuring their business impact remains considerably harder.
Jellyfish, a software engineering intelligence and AI impact platform, is addressing that gap with a new set of capabilities designed to measure how developers and autonomous agents are changing the software development lifecycle.
Announced during the company’s inaugural AI Impact Week, the features combine signals from across the engineering stack to provide a view of human development, AI-assisted work and autonomous agent activity.
The shift matters because a higher number of AI-generated lines of code does not necessarily translate into faster software delivery. If engineering teams generate code more quickly but spend more time reviewing, testing or fixing it, the productivity equation can look very different.
Jellyfish’s new Lifecycle Explorer is intended to expose those bottlenecks by showing where engineering time is being spent across the AI-driven development lifecycle.
From AI adoption to engineering impact
The platform now includes AI Cohorts, which segment developer interaction with coding assistants such as GitHub Copilot, Cursor and Claude Code.
Instead of treating AI adoption as a single organization-wide figure, engineering leaders can use the segmentation to examine differences between teams and understand how AI usage correlates with engineering outcomes.
Jellyfish is also introducing Metrics Explorer, which analyzes human contributors alongside autonomous agents. Engineering leaders can create custom metrics using natural-language descriptions, allowing organizations to define measurements around their own development processes.
An AI-powered Jellyfish Assistant and Agents capability adds a conversational layer to that data. The company says the system can surface insights, answer questions and generate views based on an organization’s engineering context.
That approach reflects a broader movement toward AI-native analytics. Rather than requiring engineering leaders to manually assemble information from issue trackers, code repositories and AI tools, platforms are increasingly expected to interpret those signals directly.
Daxko VP of Engineering Bill Pawlikowski said his team can combine information from Jira, Cursor and GitLab and query the resulting engineering data ecosystem directly.
Measuring the shift to agentic development
The new capabilities also address a more consequential change: the transition from AI-assisted developers toward autonomous coding agents.
Jellyfish’s Skill Adoption feature tracks which AI skills and practices are spreading through engineering organizations and identifies teams adopting them more quickly or slowly.
Its Behavioral Metrics are designed to measure how effectively engineers work alongside AI agents.
That distinction could become important as agentic development expands. Traditional engineering metrics were largely designed around human contributors, making it difficult to assess workflows where an agent can independently perform parts of coding, testing, investigation or remediation.
The challenge is not simply determining how much work an agent performs. Organizations also need to understand whether that activity improves delivery, creates additional review requirements or changes the skills and capacity required from engineering teams.
Jellyfish SVP of Engineering Adam Ferrari said the company has rebuilt its platform around an AI-native data foundation to address this shift.
Tracking AI spending alongside output
Cost is another part of the equation.
AI coding platforms increasingly use different pricing structures, while organizations may purchase tools directly, consume models through APIs or pay through enterprise agreements. That makes it difficult to establish a single view of AI expenditure.
Jellyfish’s new Token Usage and Spend capabilities track consumption by tool and model. Its platform also provides reconciliation between API-reported and telemetry-reported costs.
More importantly, Spend-to-work Attribution connects AI expenditure with initiatives, deliverables and roadmap activity.
The company is also introducing AI Cost Benchmarks, allowing organizations to compare spending and efficiency against other companies using Jellyfish, while Total R&D Cost places AI expenditure alongside people costs.
This could give engineering leaders a more useful measure of AI economics than tool-level spending alone. An inexpensive coding assistant may still represent poor value if it produces limited measurable improvements, while higher model costs could potentially be justified when they contribute to greater engineering capacity.
AI capacity becomes a planning metric
Jellyfish is also introducing AI Capacity, which measures changes in engineering capacity using AI while normalizing output against headcount.
That metric reflects a growing question for CIOs and engineering leaders: whether AI should primarily reduce development costs, allow existing teams to deliver more software or change how organizations plan future hiring.
The answer will likely vary by team and workload. AI can accelerate coding, but software delivery also depends on architecture, security, testing, product requirements, code review and deployment.
Jellyfish’s platform is therefore positioning AI measurement as a broader engineering management problem rather than simply a coding-tool adoption exercise.
The company’s expansion comes as enterprises increasingly experiment with AI agents and coding assistants across software development. The next phase of adoption will require organizations to establish whether those systems are generating measurable business value.
For engineering leaders, that means tracking not only how much AI is being used, but where it changes workflows, what it costs, how humans and agents collaborate and whether the resulting capacity translates into better software delivery.
Market Landscape
Jellyfish is competing in the emerging AI impact and software engineering intelligence category, where organizations are moving beyond basic AI adoption dashboards.
The competitive focus is shifting toward connecting AI usage with engineering velocity, quality, cost and capacity. Tools such as GitHub Copilot, Cursor and Claude Code provide AI development capabilities, while platforms such as Jellyfish seek to measure the organizational effects of those technologies.
The rise of autonomous coding agents makes this measurement layer increasingly important because conventional developer productivity metrics were designed primarily around human workflows.
Top Insights
- AI adoption is moving beyond usage counts, with engineering leaders increasingly seeking evidence that coding assistants and agents improve measurable delivery outcomes.
- Jellyfish connects AI spend with engineering work, allowing organizations to examine token consumption, model costs, initiatives, deliverables and total R&D expenditure.
- Agentic development requires new metrics, particularly for measuring collaboration between human engineers and autonomous software development agents.
- AI capacity could influence workforce planning, by normalizing engineering output against headcount and providing another way to evaluate AI-enabled productivity.
- Engineering intelligence is becoming an AI management layer, connecting development telemetry with financial, operational and strategic decisions.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
