Stop Buying Hours: How to Choose a Software Development Partner in the AI Era

For two decades, buying custom software meant buying time. A partner proposed a team, a rate and an estimate, and the buyer managed the risk that the hours would run over. The model was imperfect but understood.

AI has broken its central assumption: that hours are a reasonable proxy for output. When nearly every developer uses AI tools (90% of the almost 5,000 professionals surveyed in Google's 2025 DORA report), the same hour can produce very different amounts of working software depending on the team, its practices and its incentives. Paying for hours now means paying for the input whose relationship to value is least stable.

The market is responding. IDC predicted in October 2025 that by 2029, driven by agentic AI, 30% of all contracts with service providers will be outcome-based. Reuters reported in August 2026 that India's largest IT firms are increasingly tying fees to outcomes rather than billable hours. The question for buyers is how to evaluate partners, and structure contracts, for a world where effort is no longer the product.

The short answer

How should companies choose a software development partner now that AI is part of development? Evaluate partners on what they deliver, not on rates and headcount: time to a first working release in production, delivery stability, ownership of code and IP, and a measurable business outcome. Ask for evidence of engineering discipline (automated testing, deployment metrics), because AI amplifies both good and weak practices. Use time-and-materials only for genuinely exploratory work, and move defined work to fixed-scope or outcome-linked terms.

What AI actually does to developer productivity

The evidence is real but more mixed than vendor marketing suggests, and the mix matters for how contracts should be written.

Three conclusions follow for buyers. First, AI can speed up well-defined work significantly. Second, developers' own sense of their speed is unreliable, as METR's 2025 participants showed. Third, without strong testing and release practices, AI-driven speed creates instability downstream, which DORA's authors attribute to larger volumes of change exposing weaknesses in the delivery system.

In short, the gains are captured by teams with engineering discipline, and they show up in delivery outcomes, not in hours logged.

Why hourly billing now works against the buyer

Under time-and-materials, a partner that uses AI well delivers more per hour but bills fewer hours for the same scope, so its revenue falls as it improves. A partner that uses AI poorly bills the same hours for work of uneven quality. Neither incentive favors the buyer.

The largest service providers are already restructuring. William Blair's January 2026 equity research reports that 60% of Accenture's fiscal 2025 revenue was fixed-price, and that outcome-based contracts made up 47% of Cognizant's sales in the third quarter of 2025. Reuters reported in August 2026 that TCS's chief executive put the outcome-based share at about 80% of contracts in one of its segments, and that clients are asking for the same work at 25–30% lower cost.

The shift is real but early, and not universal. Forrester analyst Biswajeet Mahapatra, quoted by Outlook Business in September 2025, noted that time-and-materials still dominates because many projects are unpredictable and value is hard to define. That is a fair caution, and it shapes the framework below.

Matching the contract model to the work

Editorial framework by DaCodes. Note that fixed price is not the same as outcome-based: a fixed-price project can still deliver a system nobody uses.

How to evaluate a partner: the outcome scorecard

1. Time to first working release

Ask how quickly the partner commits to putting a working, usable piece of the system in production, and what it includes. A partner confident in its process should commit to weeks, not quarters, for a first module.

2. Delivery performance you can verify

DORA's software delivery metrics are the closest thing the industry has to a standard. Its current model groups them into throughput (change lead time, deployment frequency, failed-deployment recovery time) and instability (change fail rate, deployment rework rate). Ask partners which of these they report to clients and to show examples.

3. How AI is used, and governed, in their delivery

Ask which AI tools the team uses, how AI-generated code is reviewed and tested, and how they protect your code and data. Given DORA's stability finding, automated testing and review practices matter more than the tools themselves.

4. Ownership and exit

Confirm that you own the code, infrastructure definitions, documentation and any AI prompts or evaluation sets from day one, and that the system can be run by your team or another partner.

5. Proximity to the business

The best partners work alongside the people who use the system, not only through tickets. For US companies, nearshore teams in Latin America offer real-time collaboration in overlapping time zones. Deloitte notes that nearshoring lends itself to real-time collaboration and reduces time-zone fatigue, and, citing ILO, OECD and Gartner data from 2021–22, that Brazil and Mexico together have more than 2.2 million software engineering professionals.

A perspective that challenges the consensus

The common reaction to AI productivity claims is to negotiate lower hourly rates. That captures little value and often lowers quality, because it rewards the cheapest hour rather than the fastest result. The more effective move is to stop negotiating the hour at all and negotiate the outcome: what ships, when, at what quality, and who owns it.

Editorial note: this is DaCodes' view, informed by the pricing shift documented above.

What this means for business leaders

1. Hours have become a weak proxy for value. With AI in nearly every developer's workflow, output per hour varies widely by team and practice.
2. Engineering discipline is now the differentiator. DORA's data suggests AI amplifies the strengths and weaknesses of a delivery system; testing and release practices decide whether speed becomes value or rework.
3. Contract structure is a strategic choice. Leading service providers are moving to fixed-price and outcome-based models. Buyers who keep defaulting to hourly billing leave the AI productivity gain on the table.

Your next 90 days

1. Days 1–15: List your active and planned development engagements. Classify each by the contract table above and flag those billed by the hour for well-defined work.
2. Days 16–45: Ask current and prospective partners for the outcome scorecard: committed time to first release, DORA-style delivery metrics, AI governance, ownership terms.
3. Days 46–75: Run one defined module as a fixed-scope or outcome-linked engagement with a measured baseline.
4. Days 76–90: Compare delivery and quality against an hourly engagement of similar size, and use the result to set your sourcing policy for 2027.

Frequently asked questions

Does AI make software developers more productive? Often, but not uniformly. A GitHub experiment found 55% faster task completion, while a 2025 METR randomized trial found experienced developers 19% slower on familiar codebases. METR's 2026 follow-up suggests newer tools likely help, with weak evidence.

What is outcome-based pricing in software development? A contract in which part or all of the fee depends on delivering an agreed result, such as a working release or a business metric, rather than on hours worked. IDC predicts 30% of service-provider contracts will be outcome-based by 2029.

What are DORA metrics? DORA's software delivery metrics measure throughput (change lead time, deployment frequency, failed-deployment recovery time) and instability (change fail rate, deployment rework rate). They are a widely used standard for comparing delivery performance.

When is time-and-materials still the right model? For exploratory work where scope is genuinely unknown, such as discovery or research spikes. It should be short and time-boxed, with working software shown frequently.

Why consider a nearshore partner in Latin America? Nearshore teams work in time zones that overlap with the US, which enables real-time collaboration. Deloitte cites data showing Brazil and Mexico together have more than 2.2 million software engineering professionals.

Conclusion

The decision is no longer which vendor offers the lowest rate but which partner will commit to outcomes and prove them. Define what must ship and when, ask for verifiable delivery metrics and clear ownership, and reserve hourly billing for work that is truly exploratory. The partners worth hiring in the AI era are the ones willing to be measured by what they deliver.

Want to test an outcome-based engagement? DaCodes commits to a first functional module in production in 2–6 weeks, on terms you can measure. → Talk to DaCodes

Sources 

• Google Cloud, Announcing the 2025 DORA Report, Sep 23, 2025.
• DORA, Software delivery performance metrics.
• GitHub, Quantifying GitHub Copilot's impact on developer productivity, Sep 7, 2022 (updated May 21, 2024).
• METR, Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, Jul 10, 2025.
• METR, Uplift update, Feb 24, 2026.
• Stack Overflow, 2025 Developer Survey: AI.
• IDC, Charting the agentic future of IT services, Oct 30, 2025.
• Reuters via Zawya, AI reshapes India's IT services sector contracts, Aug 21, 2026.
• William Blair, Entering 2026: AI pricing takes hold, Jan 23, 2026.
• Outlook Business, AI pushes IT services to outcome-based deals, Sep 17, 2025 (Forrester quote is secondhand).
• Deloitte, Nearshoring as an offshore alternative (no publication date on page; talent data from 2021–22).