David Macias

New Five9 survey: 46% of customers rated their last chatbot experience positively, vs about two-thirds for a human. The top reason people bail on the bot? It repeatedly misunderstood them. Turns out the human exit isn't the failure mode, it's the feature.

Remember when every vendor said AI's focus was not to reduce agent count? Well how about it reducing your agent license count? Zoom CX is moving to one spend pool for seats, apps, and AI. A seat you can count. A draw-down pool is their meter, and renewals just got harder to justify.

A VC firm read 67 SaaS earnings calls and concluded AI is great for SaaS valuations. Anyway, the underlying numbers on outcome pricing and MCP growth are worth your time if you read past the pitch.

New Pebble watchface from me, @ericmigi.com: Multi Timezone with Date Plus. Big local time, 6 world zones, day-offset markers, wrist flick to zoom each zone. Link in first reply.

AI-agent caller detection (Sierra's fleming-1, Pindrop's BotStopper) should be native to CCaaS, not a bolt-on. It's the new caller ID: the answer changes how you route the call, so it belongs in the media path.

Gladly's outcome-based pricing sounds great until you notice: no published prices, the vendor defines the meter, and transcription/summaries that shipped in your seat license are now per-use charges. Same bill, better story.

ServiceNow's Flow is $2 per conversation, not per resolution. You pay when the exchange happens, not when the issue gets fixed. And the boundary of a "conversation" is drawn by the system billing you for it.

CXM's autonomous contact center piece has a brutally honest intro, then pivots to a vendor-style rollout vision. My take: without data governance and documented agent workflows, you're not building autonomy, you're building confident wrongness at scale.

Read the Walmart CEO's pricing letter. He promises the AI won't raise YOUR price or hide cheaper options. He doesn't promise it won't steer you to the expensive one. Recommendations are carved out.

Per-seat SaaS is dead. AI agents killed it. A certain four-letter SaaS company has new Christmas pricing (see attached). My favorite line: the meter is free (unaudited, trust us).

An AI shopping agent has no fiduciary duty. A broker who steered you to kickback funds would be breaking the law. The bot is just "optimizing." We need decision receipts: what it considered, what it rejected, and why.

Amazon joined the AI safety debate: ship models only when "safe." Great. Nobody has defined safe, and Amazon sells the cloud, the models, and the marketplace. They get to help write the definition. What could go wrong.

CallMiner's CX report: in-house AI went 6% to 17% in a year, while belief that AI "unlocks employee potential" fell 96% to 85%. Coincidence? Maybe. But nothing cools enthusiasm like maintaining the thing yourself.

I'm finally reading The Forever War by @joehaldeman.bsky.social. Yes, I'm behind the times. I'm only half way through and it's just so depressing. I could only read this during vacation. Great book so far.