PP&A designed an interview guide in eight sections: the interviewee's experience with AI agents, where in the lifecycle agents add the most value and how that value is quantified, pricing models and monetization, decision-making and risk management, commercialization and market dynamics across the three tiers of vendors, multi-agent approaches and integration formats, and expectations for the next one to two years.

In March and April 2025 the team led in-depth interviews with engineering and technology-procurement leaders who had used or evaluated agents for coding, documentation, testing or deployment. PP&A synthesized the conversations into an interview-findings readout organized around four themes: agents in the development lifecycle, budgeting and productivity, costs and pricing, and adoption and future trends.

The findings tempered the pace the client might have assumed. Organizations were actively exploring how agents could speed up software development, with interest in both cost reduction and efficiency, but traditional systems and human oversight remained firmly in place. Existing budgeting methods such as story-point systems offered no standard way to measure an agent's contribution, and team size, task complexity and tooling all shaped how work was assigned and costed, with managers still in control of both. One leader reported a twofold productivity gain for software engineering, a signal of the upside once agents are embedded.

On commercial terms, interviewees expected an early spending surge as companies set up agent systems tailored to their needs, followed by development budgets falling by about 20 percent within one to two years. Despite curiosity about outcome-based pricing, large companies preferred fixed-price contracts with set limits, accepting usage-based elements only once an agent had proven its value. A five-figure monthly subscription for a single agent was dismissed as impossible to justify, a reminder that price must track demonstrable value.

Adoption would proceed step by step through pilots and internal tests, with the coding assistants already in place setting the timing for new agent features. Leaders expected agents to become more specialized across the lifecycle and saw the real work ahead as balancing task replacement against process improvement, without major shifts in organizational structure in the near term. The readout gave the client a grounded view of pricing tolerance and adoption pace to feed its commercialization plan.

• --