Automation: Make adds fallback AI and event alerts, UiPath grades agents A to F

This week's automation news shares one goal: keeping AI agents reliable when something goes wrong. Make gives agents a fallback AI and reports failures in real time, UiPath adds a quality review for agents, and n8n supports a new Google model from day one. Here are the details.

Make: fallback provider and caching for agents

On 9 October Make shipped several improvements to the Make AI Agent app. The most important is the fallback connection: if the primary AI provider or model is unavailable, Make retries the run once on a backup connection. Make's own AI Provider is the default fallback. The Reasoning tab shows when the fallback took over.

Prompt caching comes with it. The agent reuses stable content such as instructions and tool definitions, which cuts latency and credit usage. For Claude models you choose the cache duration; for supported OpenAI models caching happens automatically. Both are available on all plans.

The same update brings three more changes: the Maia assistant reviews a scenario run and suggests how to improve the agent's response, admins can query audit logs through MCP (read-only), and social sign-ins with a company email are now routed into the organisation's SSO flow.

Make: real-time event notifications

A day earlier, on 8 October, Make introduced event subscriptions. You subscribe to platform events, such as a scenario being created or a run failing, and Make posts them to a URL you choose. Every delivery is signed with a secret, so your system can verify it came from Make. A delivery log keeps the status, response code and payload of each attempt. It is available on the Teams plan (up to three subscriptions per organisation) and on Enterprise (unlimited).

UiPath: agents get an A to F grade

On 8 October UiPath released agent review with Autopilot in preview. It replaces the previous Agent score. Autopilot finds critical issues, warnings and improvements, assigns a grade from A to F based on how many issues there are and how severe, and can fix them in the same conversation. It applies to agents in Automation Cloud and Test Cloud.

n8n: Gemini 4 Argon from day one

On 13 October n8n announced that Google's new Gemini 4 Argon model is available in the Google Gemini node and as a model for Agents. On n8n Cloud it runs on Gateway credits, with no separate provider account. Self-hosted users add their own API key. The n8n Assistant has no native Gemini option yet and connects through OpenRouter or Google's OpenAI-compatible endpoint.

What it means for your business

1. Give every agent a fallback. If a process depends on a single AI provider, an outage stops the work. In Make, a fallback connection takes minutes to set up.
2. Let failures report themselves. Instead of checking run history by hand, send failed-run events straight to your ticketing system or team chat.
3. Review agents before they go live. A grade with a list of issues, as UiPath now provides, is a good standard for agents on any platform.

If you want your workflows to keep running when something fails, see our automation service or get in touch.