Weekly signal

This week (June 22–30, 2026) the creative industries saw agentic AI move from research and pilots into product and partnership playbooks. Three developments matter: Adobe doubled down on agentic Firefly + enterprise agent partnerships at Cannes and announced the planned acquisition of Topaz Labs; Google DeepMind signed a high‑profile R&D/investment partnership with indie studio A24 to build filmmaking tools; and OpenAI published evidence that agentic tools (Codex/agents) are already shifting how organizations — including media and ad companies — allocate creative and technical work.

What changed

1. Adobe framed Creative Cloud and Firefly as an "agentic" production layer at Cannes Lions and disclosed new agency and platform partnerships that embed Adobe agents across enterprise ad and creative workflows. This is explicitly positioned as a multi‑agent orchestration play (MCP/connectors, brand governance, platform integrations).

2. Adobe announced a definitive agreement to acquire Topaz Labs (image/video enhancement AI) to bring on‑device enhancement models and studio‑grade restoration/upscaling into Firefly and Creative Cloud. The deal is expected to close in H2 2026 and signals Adobe’s push to combine generative + capture‑preservation tooling.

3. Google DeepMind and A24 launched a multi‑year research partnership backed by a reported ~$75M investment to prototype AI tools for filmmaking (previs, storyboards, production workflows). That brings a top research lab directly into a studio product loop and signals more vertically tailored agent tooling for film.

4. OpenAI published an internal/economic research post documenting rapid adoption of Codex/agents across functions: agents are being used for longer‑horizon, cross‑functional work and non‑developer adoption is accelerating — a practical signal that agentic workflows are ready for mainstream creative teams (ad ops, production, post). OpenAI also published an HP Frontier partnership case showing enterprise agent runtimes and governance playbooks.

What to do with it

• Product leaders (studios, agencies, tooling vendors): run a 2–4 week agentic pilot that focuses on one measurable outcome (e.g., storyboard to edit‑ready shot list, or campaign package generation). Lock scope, permissions, human check‑ins, and cost caps. Prioritize connectors (MCP) to your DAM, edit NLEs (Premiere/Resolve), and asset metadata.

• Creators & post houses: evaluate on‑device quality/latency from Topaz‑style models vs cloud generative steps. Test hybrid flows where AI enhances captured footage (restoration/upscale) rather than replacing artistic decisions; preserve provenance metadata for credits and licensing.

• Legal / Vfx / IP teams: renegotiate vendor and talent agreements to clarify training‑data, derivative rights, and acceptance criteria for agent‑produced deliverables before scaling. Log agent decisions and keep clear audit trails.

• Agencies & brands: treat agentic features as operational risk + multiplier. Prototype an agentic creative operating model for one brand vertical (sports, retail) with a gated rollout and brand‑safety checks integrated into the agent’s evaluation loop.

• Builders/engineers: use this window to harden agent governance: least‑privilege tool access, execution step limits, redaction in logs, and human‑in‑the‑loop checkpoints aligned to creative sign‑offs. Use on‑device inference where latency, cost, or IP control matters.

(Primary sources and reporting cited below.)