|
Members: no offer this week. You already have everything DeskTheory makes. Straight to the news.
|
|
|
|
In AI this week
|
AI layoffs keep coming. The data says they are not where the returns are.
AI-cited job cuts kept rolling through June, with TechCrunch's tracker logging GitLab trimming about 14% of its staff. But a Gartner study of 350 large enterprises found that the companies reporting real AI ROI laid people off at the same rate as those seeing little or none. The returns came from retraining the people who use the tools, not replacing them. (Gartner, Fortune)
What it means for you: The reflex is to cut headcount and book the AI savings. The data says the payoff comes from amplifying the people you keep, not the ones you cut. If you are weighing a team reduction against retooling that team, the evidence points hard at retooling, which is the harder plan to run well.
|
|
|
KPMG just put governed AI agents on 276,000 desks
On June 9, KPMG and Microsoft said KPMG will roll Microsoft 365 Copilot and Agent 365 across its entire 276,000-person workforce in 138 countries. Agent 365 is not a chatbot; it is the governance layer that manages, monitors, and audits AI agents at scale, and Microsoft tagged KPMG a "Frontier Firm" for redesigning work around it. (Microsoft)
What it means for you: The headline is "agents," but the real move is the governance (identity, policy, audit) that made a risk-averse firm comfortable running them firm-wide. The Big Four are not debating whether to use agents; they are building the controls to scale them. The firms you hire and bid against are changing their cost structure while you read this.
|
|
|
Your AI bill is becoming a finance problem
On June 18, OpenAI gave ChatGPT Enterprise admins usage analytics and hard spend controls: a console showing who is using AI and what it costs by user, team, and model, plus per-workspace credit limits and a cost API. (OpenAI)
What it means for you: As AI spreads past the early adopters, the spend gets hard to see and harder to govern. When the vendor itself ships spend controls, it is telling you that "shadow AI" has graduated from an experiment into a line item. Get a one-page usage policy in place before the invoice forces the conversation.
|
|
|
|
This week's workflow
|
|
Roll out AI to one team in 30 days
|
|
|
WORKFLOW / THE +1
Roll out AI to one team in 30 days
> OUTCOME
One team, one workflow, one champion, measured against a baseline, in thirty days. Instead of a company-wide announcement that quietly dies, you land one real win you can prove with numbers, then repeat the template with the next team. A 2025 MIT study found roughly 95% of corporate AI pilots produced no measurable P&L impact, not because the models are weak, but because the rollouts were. This is the opposite: narrow and deep.
> SETUP
Pick one high-frequency, low-risk workflow for a single team (drafting replies, summarizing tickets, prepping reports) and baseline it first: time per task, volume, a quick quality rating. Week 2, hand them a Claude Project preloaded with tested prompts and name a respected peer, not someone from IT, as champion. Week 3 they use it on real work with the champion in the flow. Week 4, re-measure and decide: kill it or double down.
> WHAT YOU SHOULD DO NEXT
This week, pick the one team and the one workflow, then baseline it before you change anything. Measure.
|
| |
↳ Full walkthrough _
|
|
|
Andrew
|
|
1:1 · OPERATOR TO OPERATOR
|
$1,000 · limited spots
|
Want to work through any of this directly?
I do a small number of Operator to Operator sessions each week. 60 minutes, screen-share, you bring the bottleneck and we leave with a concrete plan you can run that week. $1,000, pay then book.
|
|
|