Summary of "No APIs, No AI: How Software Engineering Must Change"

High-level takeaway

To scale generative AI (GenAI) successfully you need three aligned changes: team topology, platform engineering, and a mature API strategy. Missing any of these creates cost, security, and quality risks that block production adoption.

Manjunath Bhat — Scaling GenAI with team topologies

Key problems when scaling GenAI:

Selecting high‑value use cases.
Meeting security and governance at scale.
Architecture/design tradeoffs (open vs closed models, self‑hosting vs third‑party APIs).
Cost blindness (tokens, GPUs).
Duplicated effort and inconsistent practices.

Recommended organizational model: team topologies with four team types

Stream‑aligned (product) teams — build customer‑facing functionality.
Enabling teams — internal consultancies that bring specialized expertise (e.g., a GenAI center of excellence combining IT, risk, legal, finance).
Complicated subsystem teams — own complex subsystems to reduce cognitive load.
Platform teams — provide curated, reusable components, recipes, and templatized workflows so product teams don’t reinvent patterns.

Examples and practical steps:

Example: Verizon’s GenAI CoE + platform teams producing end‑to‑end recipes for common design patterns (content generation flows).
Practical steps: inventory applications/use cases; identify scaling issues and competency gaps; institutionalize best practices so proofs of concept go into production.

Akis Sklavounakis — Platform engineering to reduce cognitive load

Problem:

Growing complexity (now increased by AI) raises developer cognitive load and hurts velocity.

Solution:

Platform engineering — build self‑service internal platforms that are optional but are the path of least resistance.

Five foundational principles for successful platform engineering:

Run the platform as a product (apply product management discipline).
Be demand‑driven (prioritize developers’ biggest pain points).
Provide compelling, easy paths for building software.
Offer self‑service functionality.
Start with a thin/viable platform and iterate.

Measure success across stakeholders:

Product teams, end users, business results (finance/ops/infrastructure), and the platform team itself.
Continuously gather developer feedback and iterate.

Shameen Pillai — APIs are mandatory for GenAI/agentic AI

“No APIs = No GenAI/agentic AI.” Poorly managed APIs put AI initiatives and organizational data at risk.

How GenAI affects API strategy and tooling:

Gateways: modern lightweight gateways plus specialized AI gateways / MCP server gateways.
API security: specialist tools for automated policy generation, design‑time security, proactive/predictive protections and threat detection.
Developer enablement: GenAI can generate API specs/implementations from natural language and speed multi‑language code generation.
Lifecycle: GenAI can prioritize APIs, automate validations/standards enforcement, and enable agentic workflows for API governance.

Prediction:

By 2028, >50% of API usage will originate from AI agents, not human developers.

Gartner API strategy maturity model (assessment + remediation)

Five dimensions:
1. Business alignment
2. Developer enablement
3. API security
4. Lifecycle management
5. API governance
Five maturity levels: initial → developing → defined → managed → optimizing

Recommended actions by dimension:

Business alignment: appoint API product managers; align KPIs to business goals.
Developer enablement: invest in developer relations, design standards, and self‑service registration.
API security: implement API discovery, adopt AI‑based security tools, and review agentic API use.
Lifecycle: deploy API management solutions, enable distributed management, use AI gateways, and automate validations.
Governance: create organization‑wide API policies, involve lines of business, and automate policy validation/enforcement.

Short action:

Run a formal maturity assessment, involve developers, API consumers, and business stakeholders, then create an action plan.

Practical guides and takeaways

Three‑pronged rollout guide: inventory → identify gaps → institutionalize best practices.
Platform engineering playbook: run the platform as a product; be demand‑driven; enable self‑service; start thin and iterate.
API maturity assessment and remediation checklist: use Gartner’s five dimensions and maturity levels.
Example pattern: platform teams provide templatized, end‑to‑end recipes (design tools, reusable components, workflows) so product teams adopt consistent, secure practices.

Main speakers / sources

Host: Karen Stokes Lockhart (Gartner ThinkCast)
Analysts: Manjunath Bhat (Gartner Distinguished VP Analyst), Akis Sklavounakis (Gartner Senior Director Analyst), Shameen Pillai (Gartner Senior Director Analyst)