Aug 9 – Sep 8, 2026 · 6 items
The AI week, for a product manager
A real edition, written for: Product Manager
The one thing
Cheap bulk transcription with speaker labels turns your interview backlog into synthesisable material this quarter.
Bulk transcription drops to a dime an hour with speaker labels
Microsoft AI released MAI-Transcribe-2, a speech-to-text model, on Thursday. The launch price is 10 cents per hour of audio, called an early-bird rate. Microsoft has not named an end date or a standard price. The model covers 60 languages and handles noisy, overlapping real-world audio. It labels who is speaking, timestamps each word and accepts custom word lists. A verbatim mode keeps filler words for legal and compliance use. It also follows conversations that switch language mid-sentence. Microsoft claims first place on the FLEURS multilingual benchmark and second on Artificial Analysis. It says the model runs five to ten times faster than rivals from OpenAI, Google and ElevenLabs. The announcement says nothing about real-time transcription, speaker-labelling accuracy or data retention.
Why it matters for you
Your discovery interviews are the raw material for research synthesis. Transcribing a full quarter's calls is now a rounding error.
- Speaker labels and timestamps mean you can quote a specific user, not paraphrase
- It runs through Microsoft Foundry, so this is a request to whoever owns your Azure account
- Sixty languages covers DACH-region interviews you currently handle by hand
Try this
Take three past interview transcripts into ChatGPT and pull the recurring pain points before asking for bulk transcription.
Paste this into your AI tool
Here are three customer interview transcripts from my B2B SaaS product area: [paste them]. Identify the problems mentioned in more than one interview. For each, quote the exact line from each transcript that supports it, name which interview it came from, and say how confident you are. Then list the things only one person said, separately, so I don't over-weight them. Do not add problems that aren't in the text.
OpenAI's Astra operates browsers and spreadsheets directly
OpenAI released GPT-6 Astra, a frontier model built to operate computers directly. It works across browsers, spreadsheets, websites and desktop applications. It can fill forms, update CRM records, run web research and produce documents. OpenAI says this reduces the need for hand-built connectors to each business system. President Greg Brockman told a press briefing that the company is now in the AGI era. Rollout starts Thursday for enterprise customers in the Daybreak gated program. Paid ChatGPT tiers, the OpenAI API, AWS Bedrock and Microsoft Azure follow in coming days. Brockman argued buyers should compare price per completed task rather than per token. OpenAI omitted GDPval, its own benchmark for real-world occupational work. OpenAI also paused some frontier training for about two weeks after the Hugging Face incident and tightened infrastructure controls.
Why it matters for you
You already run ChatGPT. This model works inside apps rather than just answering, which touches the spec and launch admin you do by hand.
- It reaches paid ChatGPT tiers in coming days, so no new tool to buy
- Enterprise access starts gated, meaning your IT will decide the timing at 251-2500 people
- Watch the scope-adherence claim: an agent editing a real backlog needs a checkpoint before it saves
Try this
List the three roadmap chores you do in a browser weekly, and ask IT when Astra reaches your tenant.
Slack Code puts coding agent work in a channel non-engineers can watch
Salesforce launched Slack Code, which gives teams dedicated channels where people and coding agents work on a task together. Tagging an agent creates a channel that shows the conversation, the code changes and a live preview, then archives itself when the work is done while keeping a searchable record. It works with agents from Anthropic, Cognition, GitHub, OpenAI and Vercel on any Slack plan, but access to each agent must be bought separately.
Why it matters for you
You own specs and launch but sit outside the repo. This makes agent-driven changes visible where you already talk to engineers.
- A live preview and line-by-line diff in-channel means you can approve before ship without a Git account
- The channel archives itself as a searchable record, useful when a launch decision gets queried later
- Each partner agent is bought separately, so this is a conversation with your eng lead, not a purchase you make
Try this
Ask your engineering lead whether any of your team's coding agents are Slack Code partners.
Asana bundled AI into existing tiers and took a revenue hit for it
Asana will bundle a base level of AI tools into its work management plans from mid-September. The tools include AI Teammates and Dash, and tier prices do not change. The change applies to both new and existing customers. Asana will also recognise revenue from new AI Teammates sales as customers use their requests. CFO Aziz Megji said the shift creates a revenue headwind and pressure on gross margins in the second half. CEO Dan Rogers said AI products made up about a quarter of new recurring revenue last quarter, up from the prior quarter. Asana raised its full-year target for AI's share of new recurring revenue. It also signed its largest AI expansion deal, a three-year agreement with a Fortune 500 media company. Both executives spoke on an earnings call on Thursday.
Why it matters for you
You own pricing-adjacent decisions for a B2B SaaS area. A peer just chose bundled AI over an upcharge and published the cost.
- Bundling to both new and existing customers is the choice most PMs are debating right now
- Usage-based recognition on AI features changes when revenue lands, which your finance partner will care about
- AI share of new recurring revenue rising quarter on quarter is the number that justified the hit
Try this
Draft the two-option memo — bundle versus upcharge — for AI features in your product area.
Paste this into your AI tool
I am a product manager for a B2B SaaS product area. I am deciding whether to bundle new AI features into existing pricing tiers or charge separately for them. Here is my product area and current tier structure: [describe it in a few lines]. Write a one-page memo with both options. For each, give the argument for it, the strongest argument against it, what it does to revenue timing, and what would have to be true for it to be the right call. End with the three questions I should put to finance before deciding. Be concrete, no generic pricing theory.
Sources
Meta cancelled a layoff round after AI agents underdelivered
Meta cancelled the second wave of layoffs planned under an internal reorganisation that would have shifted much of the daily work of thousands of employees onto AI systems overseen by small expert teams. The company still cut about a tenth of its staff in May, but internal measures showed autonomous AI agents were not delivering the expected productivity, while technical and security incidents rose. Zuckerberg told staff in July that the pace of AI agent progress had been misjudged.
Why it matters for you
You are being asked to put agents on your roadmap. This is the clearest counter-evidence available, with numbers attached.
- Code volume rose sharply while incidents and fix times rose faster — throughput is not the outcome metric
- Developer sentiment fell, which matters if agents touch your own team's workflow
- Useful ammunition when a stakeholder proposes agent automation without a quality gate
Try this
Add an incident-rate and rework metric to any agent feature already on your roadmap.
Paste this into your AI tool
I am a product manager for a B2B SaaS area. I have an AI agent feature on my roadmap: [describe it in two lines]. Propose five success metrics that would catch it going wrong, not just metrics that show it being used. For each, say what a bad reading looks like and how soon after launch I would see it. Then name the one metric I should put in the spec as a rollback trigger.
Sources
Only 22% of large firms have scaled AI beyond one business unit
Gartner found that only 22% of large organisations have scaled AI across several business units. It surveyed more than 1,300 leaders at firms with over $50 million in revenue, between January and April. Spending plans are undented, with 85% of technology leaders raising AI budgets next year. About 11% of respondents could not say what they spent on AI in 2025. Gartner's Tina Nunno warned that weak measurement tied to business outcomes wastes resources. Firms that track returns continuously and shut down weak projects reported gains on 81% of initiatives. Popular uses such as cybersecurity, threat detection and IT service desk automation often return less. The best returns came from IT asset and cost optimisation, synthetic data generation, and automated code generation. Separate reports from Infosys and Deloitte found similar gaps in measurement and readiness.
Why it matters for you
You prioritise a roadmap against other bets. This says which AI categories actually returned and which quietly did not.
- Firms that tracked returns continuously and killed weak projects reported gains on most initiatives
- Automated code generation and cost optimisation returned better than the popular security use cases
- A sizeable share of respondents could not state last year's AI spend, which is the measurement gap to avoid copying
Try this
Write the kill criteria for each AI item on your roadmap before the next planning round.
Paste this into your AI tool
Here are the AI-related items on my product roadmap: [list them with one line each]. For each, write a kill criterion: a specific, measurable condition that, if true 90 days after launch, means we stop investing. Make each one falsifiable and tied to a business outcome, not usage. Flag any item where I have not given you enough to write a real criterion, and say what I need to decide first.
Build this
Every week, one small thing to build with AI in something you actually care about. No work in it. Five minutes to set up, and worth keeping if it earns a second run.
A finished autumn hiking plan built only from the Swiss trails and huts you have already walked and already booked.
Five minutes to set up
I hike in Switzerland. Here is what I have already done this season, with rough dates and how each felt: [list the trails, huts or day walks you have already done]. Here is what I already have booked or already own for the rest of autumn: [list bookings, passes, or free weekends]. Build me a concrete plan for the remaining weekends using only these. Repeat routes are fine — say what to do differently on the second run. Do not suggest anything that needs a new booking, a new permit or new kit. Give me the order, the reason for each, and one route to drop if the weather turns.
- Have your season's walks and any existing bookings to hand before you start, in rough dates.
- Run it in ChatGPT so you can push back in the same thread.
- Check any hut or transport timing it states — assume it guessed and verify before you rely on it.
- If a weekend in the plan looks wrong, tell it which one and why, and have it rebuild only that part.
Yours arrives Thursday.
This one was written for a product manager. Tell us what you do and the next one is written for you — same news, your job, once a week.