Key takeaways
- Google's agents run under their own accounts with an audit trail. That tells you where enterprise agent design is heading: identity, permissions and logging come first.
- Platform agents suit tasks inside one vendor's ecosystem. Custom agents suit workflows that cross your own systems and rules.
- Gartner warns that over 40% of agentic AI projects may be cancelled by the end of 2027, mostly because the value is unclear or risk controls are weak.
- Start with one measurable workflow, approved tools only, and a human checkpoint before anything can't be undone.
What Google announced
According to TechCrunch’s report of 8 October 2026, Google is bringing agentic AI to Gemini, starting with businesses. Here’s what matters if you run one:
- Each agent gets its own Workspace account, so what it does can be traced and controlled like any other user.
- Every action goes into an audit trail.
- Agents can connect to Google Workspace, Microsoft 365, Slack, Jira, BigQuery and MCP servers.
- Companies can choose the underlying model, including Anthropic’s Claude, and set spend caps.
The report didn’t include pricing or a general availability date, and we haven’t tested the product. Treat it as a sign of where things are going, not something to budget around today.
Why it matters even if you never use Gemini
It’s one more sign that AI agents are turning into an ordinary product category instead of an experiment. Three design choices stand out, and they apply to any agent you build or buy:
- Identity. An agent working in your systems should have its own credentials, not borrow a person’s login.
- Least privilege. It should only reach the tools and data its job needs.
- Traceability. Every action gets logged so a person can review it later.
We use the same principles in our AI agent development projects. They’re what separates a useful assistant from a liability. If the term is new to you, our guide to agentic AI consulting covers the basics.
Platform agents or custom agents?
Big-vendor agents will be handy for work that stays inside their ecosystem, like drafting documents or summarising email. Custom agents make more sense when work crosses systems, follows your own rules or touches customer data. That’s our view, not a vendor claim:
| Platform agent | Custom agent | |
|---|---|---|
| Best for | Common tasks inside one suite | Workflows across your own CRM, ERP, website and inbox |
| Setup | Configure and connect | Design, build, test and integrate |
| Control | Within the vendor’s options | Rules, tone, approvals and data handling set by you |
| Risk to watch | Lock-in and generic behaviour | Cost and maintenance if scope grows unchecked |
Plenty of businesses will end up using both. What decides it is rarely the technology. It’s whether the workflow is well defined. Our comparison of AI agents and workflow automation helps you check that first.
A note of caution from Gartner
Back in June 2025, Gartner predicted that over 40% of agentic AI projects will be cancelled by the end of 2027. It blamed rising costs, unclear business value and weak risk controls. It also called out “agent washing”, where existing chatbots and automation tools get relabelled as agents, and estimated that only about 130 of the many agentic AI vendors are the real thing.
That’s not a reason to stay away from agents. It’s a reason to pick use cases where an agent clearly beats simple automation, and to prove the value on a small scope first.
Five questions to answer before you deploy any agent
- What exact job is it doing? Write it as one sentence with a start and an end.
- Which tools and data can it touch? List them. Everything else stays off-limits.
- What can it do without asking? Anything irreversible, like sending money or deleting records, needs a person’s approval.
- How will you know it’s working? Choose a measure, such as response time or hours of manual work, before launch.
- Who reviews its mistakes? Name a person and a weekly time slot.
What to do this quarter
Pick one repetitive workflow, like qualifying inbound inquiries or preparing weekly reports, and run a small pilot using the five answers above. If you’d like help choosing and scoping it, our AI strategy consulting maps the opportunities and keeps expectations realistic, and the cost guide shows what pilots usually involve. When you’re ready, book a strategy call.
Frequently asked questions
What did Google announce for Gemini in October 2026?
Do small businesses need custom AI agents if big vendors offer them?
Is it safe to give an AI agent access to business systems?
How should a company start with agentic AI?
Sources
- TechCrunch: Google brings agentic AI to Gemini, starting with businesses (8 Oct 2026)
- Gartner: Over 40% of agentic AI projects will be cancelled by end of 2027