Platform teams that predominantly build features from scratch are roughly 3.5 times more likely to report lacking a clear strategy than those who buy tools or build from templates, according to the 2026 Future of Platform Engineering report published by Octopus. The survey of 379 platform professionals reveals that while AI agents have made custom development faster than ever, the choice between building and buying still exposes fundamental gaps in platform planning. Teams building from scratch are less likely to meet their platform goals, the report finds, even as coding tools eliminate much of the effort once required to write software.
Across all features that platform teams support, purchasing an underlying tool is the dominant strategy, chosen by 51.3% of respondents as their primary method, compared to 31.2% who build on existing templates and just 17.5% who build from scratch. High-performing teams lean even more heavily on commercial tools, with 58.8% preferring them over templates (47.8%) or custom builds (33.3%). Among teams classified by their dominant build method, 57.9% of those building from scratch cite lack of clear strategy as an obstacle, versus only 15.6% of template builders and 16.7% of tool buyers. When platform teams do choose to build, the most common justification is cost-benefit analysis (29.6%), followed by a preference for open source (28.6%), fit-and-requirements assessment (22.4%), preference for commercial tools (22.4%), and building for core differentiation (16.3%). Teams with ephemeral environments on their platforms show a strong correlation with positive AI impact: more than three-quarters of teams reporting an extremely positive effect from AI on delivery speed and stability had ephemeral environments available.
One respondent captured the prevailing logic: "Build what makes us unique, buy or adopt what's standard, and always consider long-term cost and lock-in." The report notes this mirrors the principle from domain-driven design, which separates the core domain—where custom work creates competitive advantage—from supporting and generic domains, where bespoke solutions mostly add complexity and maintenance overhead for capabilities that could be purchased. Teams building from scratch are also more likely to use AI moderately or extensively for legacy modernization and language migration, suggesting they're leaning on agents to handle the maintenance burden that follows custom features. Certain features naturally skew toward templates rather than tools, including local development environments, one-click projects, and ephemeral environments, because they're tightly coupled to specific tech stacks and golden paths that off-the-shelf products rarely accommodate.
The report argues that AI has sharpened rather than simplified the strategic question. Agents eliminate the effort of writing code but not the work of owning it—maintenance, upgrades, and governance remain while the volume of change moving through the pipeline increases. The research draws on work by Simon Wardley, noting that teams unable to see what tools already exist tend to build their own solutions to problems that have been solved better elsewhere, and on DORA's finding that organizational gains from AI depend on platform quality: strong platforms amplify AI's effects, while weaker ones dampen them. The gap between having an idea and having a working tool has nearly closed, the report observes, meaning an agent can now build a custom incident-management tool before anyone has asked whether one is needed. The recommendation is clear: where a team builds from scratch, there should be a reason, and as agents write more code, the value of tools that solve well-defined problems only grows. For platform teams navigating an environment where generating software is trivial but governing it isn't, the discipline of deciding what not to build may matter more than the ability to build fast. Organizations that mistake speed for strategy risk accumulating technical debt at unprecedented scale, with maintenance costs that agents can ease but never eliminate.