1. Headline & intro
Google is quietly turning Opal from a fun “vibe-coding” toy into something much more consequential: an AI engine that can design and run workflows on its own. If you’ve ever hacked together a spreadsheet, a Zapier automation, or a Notion template to avoid writing code, this shift matters to you. What Google announced is not just another Gemini feature – it’s a step toward everyday users orchestrating small “software systems” via chat. In this piece, we’ll look at what changed, why it threatens a whole crop of no‑code startups, and what it could mean for European businesses once Opal arrives here.
2. The news in brief
According to TechCrunch, Google has added a new agent to Opal, its “vibe-coding” app that lets people create mini web apps from natural language descriptions. The agent uses the Gemini 3 Flash model to generate and execute automated workflows based on text prompts.
Instead of only scaffolding UI-style mini apps, Opal can now plan multi-step tasks, pick which tools to use, and call them autonomously. TechCrunch notes that the agent can, for example, use Google Sheets to store state across sessions, such as a persistent shopping list for an e-commerce helper. The agent is interactive: it can ask follow-up questions or present choices when it needs clarification. Google claims this lets non-technical users build comparatively complex workflows. Opal, initially launched in the U.S. in July 2025 and later expanded to more countries, is also available through the Gemini web app’s visual editor.
The report places this move in a broader wave of “natural language to app” tools from startups like Lovable, Replit, Wabi, Emergent and Rocket.new.
3. Why this matters
What Google has done here is quietly collapse three categories into one: chatbots, no‑code app builders, and workflow automation tools.
Until now, most generative AI products for non‑technical users have been stuck in one of two modes:
- Conversational helpers (ChatGPT, Gemini chat, Copilot) that answer questions but don’t remember much structure beyond the current or saved chats.
- No‑code builders (Bubble, Glide, Replit’s AI tools) that turn natural language into code or UI, but still expect users to think like software designers once things get complex.
Opal’s new agent tries to bridge that gap. You describe an outcome; the system not only sketches an interface, but also:
- Plans steps to get there.
- Decides which Google tools it needs.
- Persists and re-uses data across sessions.
- Asks you targeted questions when the plan is under-specified.
That sounds small, but it shifts power.
Who stands to benefit?
- Non‑technical workers and small teams gain a path to automation without going through IT or learning a specialist platform. Think of internal tools, approval flows, content pipelines, or lightweight e-commerce helpers.
- Google gains a wedge to pull people deeper into its ecosystem, using Sheets or other services as the “memory layer” for these agents.
Who might lose?
- No‑code startups whose main pitch is “describe your app in English” now face a heavyweight competitor shipping that promise at OS scale.
- Niche automation tools that rely on simple triggers and actions (the long tail of Zapier clones) will be under pressure if Opal can compose multi-step flows with less manual setup.
The immediate implication: the bar for what counts as a “product” rather than a “prompt” just moved. A single text description can now generate something that persists, orchestrates tools, and evolves over time. That will change how teams prototype and how quickly internal tools appear – and disappear.
4. The bigger picture
This launch fits a broader pattern: 2024–2026 is when AI systems stop being “autocomplete for text” and start acting as agents that operate software on your behalf.
We’ve seen earlier signals:
- OpenAI’s GPTs and Assistants allow custom agents that call tools and APIs, but they’re still mostly bound to chat interfaces.
- Microsoft Power Automate and Copilot Studio have been layering natural language over automation for business users inside the Microsoft 365 world.
- Legacy players like Zapier, Make and IFTTT have been inching toward AI-generated workflows, but are constrained by their bolt‑on nature.
Google, in contrast, is trying something more integrated. Opal sits close to both Gemini and the broader Workspace stack. Combined with existing products like AppSheet and Apps Script, Google is essentially experimenting with a continuum that runs from:
“Just do this once for me” → “Turn this into a reusable app” → “Make it a self-updating workflow that runs in the background.”
Historically, the tech industry has tried to democratise programming many times – from Visual Basic macros to IFTTT recipes and modern no‑code builders. Each round hit similar walls: complexity, maintainability, and the need for someone who actually understands logic.
What’s different now is that LLMs can act as the intermediate layer, translating fuzzy human intent into structured plans, and then continuously re‑interpreting those plans as situations change. Opal’s use of Sheets as a memory layer is a pragmatic move: instead of inventing a new database for non‑technical users, it leans on a tool millions already understand.
This tells us where the industry is heading:
- The UI of software creation becomes chat + light configuration.
- “Programming” shifts from writing code to managing AI agents and data boundaries.
- The competition will be less about raw model quality and more about who owns the workflow surface where users describe and run their automations.
On that last point, Google just signalled it fully intends Opal to be one of those surfaces.
5. The European / regional angle
European users may notice one thing first: Opal still hasn’t officially landed in the EU. The countries listed so far are outside the bloc. That’s not an accident; it’s an indicator of how carefully U.S. tech giants are now tiptoeing around upcoming rules like the EU AI Act, in addition to GDPR and the Digital Services Act.
An AI agent that plans and executes workflows, possibly touching personal or business data in Sheets and other tools, will raise regulatory questions in Europe:
- Under the AI Act, even “general purpose” systems need transparency about capabilities and limitations, plus technical documentation and risk management.
- Under GDPR, any automated processing of personal data – especially if used in HR, credit, or public services – triggers strict obligations around purpose limitation, lawful basis and user rights.
Expect Google to delay an EU rollout until it can offer stronger guardrails: logging of agent decisions, admin controls for organisations, and clear data‑processing terms. For European SMEs, though, the potential upside is big. Many lack dedicated developer teams but live in spreadsheets and docs; Opal‑style agents could be their first serious automation layer.
There’s also a competitive angle. Europe already has strong players in automation and process intelligence – from Czech‑born Make (formerly Integromat) to Germany’s Celonis in process mining. As Opal matures, these companies will need to decide whether to integrate tightly with U.S. foundation models, double down on privacy‑centric offerings, or both.
For now, European users are mostly watching from the sidelines – but they should be preparing internal policies and skills for when similar agentic tools inevitably appear in locally compliant form.
6. Looking ahead
Several things are worth watching over the next 12–24 months.
Where Opal lives inside Google’s universe. Right now it’s a semi‑independent app plus a Gemini web integration. If Google is serious, expect tighter links to Workspace (Gmail, Docs, Sheets) and possibly Android. The more surfaces it reaches, the more it competes with entrenched no‑code and RPA tools.
Governance and security features. Once non‑developers can spin up workflows that touch business data, companies will demand:
- Role‑based access controls for what agents can do.
- Clear logs of every action an agent takes.
- Easy ways to audit or roll back changes.
Business model. Today, tools like this often hide inside broader AI subscriptions. Over time, expect:
- Tiered pricing for heavier automation usage.
- A marketplace of reusable Opal “recipes” created by agencies and freelancers.
Impact on jobs and skills. Opal will not erase software engineering, but it will erode some of the “glue work” developers do: wiring forms, spreadsheets and APIs together. New roles will emerge around prompt systems design, AI policy, and oversight rather than raw coding.
The big open question is how reliable these agents can become. LLM‑driven planning is still prone to hallucinations and edge‑case failures. In a chat, that’s annoying; in an automated workflow touching invoices or inventory, it’s material risk. The winners in this space will be those who pair flexibility with strong guarantees and observable behaviour.
7. The bottom line
Google’s new Opal agent is more than a clever demo; it’s a visible step toward everyday users orchestrating software via conversation. That raises the stakes for no‑code startups, pressures incumbents in automation, and intensifies Europe’s regulatory puzzle around AI agents. The key question for readers is not whether this wave is coming, but how ready your organisation is to let non‑developers design – and govern – their own AI‑driven workflows.
