OpenAI’s Dot and Anthropic’s Duet each found one genuinely useful item during a ZDNET writer’s three-week test, but most alerts were irrelevant or stale. The review captures the trade-off at the heart of proactive AI assistants: a timely catch can help, but monitoring your accounts does not guarantee the assistant will know what matters.
Watch Desk analysis
What happened
Dot and Duet monitor connected services and push items they judge worth attention. In ZDNET’s account of the test, Duet spotted a Chase chargeback notice that the writer’s existing support filters had missed by two days. The writer says they followed up and resolved the issue.
That was the standout. The review also describes alerts about an old customer-support issue and an announcement from the previous week. The writer says both assistants missed other important messages, while available connectors and reluctance from a client to grant access to Asana limited what the tools could monitor. AI processing counts against usage allocations, the review notes, so constant checking may use credits without producing a constant stream of useful insight.
Why it matters
These assistants need access to personal or workplace accounts to spot things users have not asked about. Yet more access does not automatically mean better judgement: a tool can see a busy inbox and still elevate the wrong message. And where a service is not connected, or an organisation blocks access, the assistant’s view is simply incomplete.
The review’s example is unusually clear about the upside and the catch. One missed chargeback mattered enough to justify an alert; stale and irrelevant notifications made up much of the rest. That is a useful distinction for anyone weighing proactive help against the permissions and usage costs it asks for.
Our read
The chargeback catch shows why this idea is tempting. The rest of the test shows why “it monitors your accounts” is not the same as “it understands your priorities”. Before connecting one to work or personal services, check which connectors you actually use, what access they require and how AI processing affects your allowance. A clever alert is welcome; an expensive stream of digital shoulder-tapping is less so.
What to watch
- Whether Dot and Duet improve the relevance of alerts over longer use.
- Which connectors they add, and what level of access each requires.
- How much ongoing monitoring uses from paid usage allowances.
Discussion spark: Would you grant an AI assistant access to more of your accounts for the chance of catching one important alert, or is the noise and access cost too high?
Sources and evidence
- I tested OpenAI Dot and Anthropic Duet for weeks – and all I got was one helpful notification – ZDNET (9 November 2026, 13:24 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.