Three signals this month, three reasons to pay attention.
First, Anthropic just confirmed its models write over 80% of merged production code. Marlene digs into the real story: not raw productivity, but who is actually steering the ship, and which engineering skills still matter when the humans mostly approve.
Then Slavena flips the script on enterprise buying. When AI agents join the committee, technology selection bends toward what machines can discover, trust, and act on.
Finally, Jon follows the money like Bobby Axelrod on a 3AM market play. Uber burned its entire 2026 AI coding budget in four months, and Microsoft hit the brakes too.
The engineer who hasn't written code in five months
Anthropic just confirmed their models write over 80% of merged production code, with engineers shipping 8× more per quarter. The Guardian picked it up and so did everyone else. But most coverage missed the real story. This isn't only about productivity gains but more about who is in control. Anthropic's call for a coordinated pause comes from their lab watching its own data and asking whether the humans are still steering or just approving. The same question applies to every builder right now. What are the skills really matter at the moment?
Stay tuned for our next episode with Tom Shaw where we explore how the skill set of engineers has changed.
For decades, enterprise software purchasing has focused on human users. AI agents are about to influence those decisions as key consumers of enterprise systems and, who knows, perhaps one day as buyers. Their need for structured data, predictability, explainability, and machine-readable outputs is bound to reshape technology selection.
The more interesting shift is for human decision-makers. Rather than evaluating every tool directly, they will define the guardrails: security, compliance, approved vendors, budgets, and governance. The question morphs from whether people can use a system effectively to whether agents can discover it, understand it, trust it, and act on it safely.
Scaling AI isn't just about speed. It's also about token consumption.
In 4 months, Uber blew through its entire 2026 AI coding budget. And just 6 months after telling employees they could use Claude Code, Microsoft is winding down the experiment. Axios reports that another company spent half a billion dollars in a single month after failing to impose usage limits on employee Claude licenses.
Without proper governance and value management, the unit economics of enterprise AI coding at current token prices do not work.
Globalization expanded our exposure to new ideas. Generative AI may be nudging us the other way.
Researchers call this generative monoculture: leading AI models are trained on similar datasets, often dominated by English-language content, causing them to produce similar answers and viewpoints.
One simple workaround: change the language of your prompt. Asking the same question in English, Chinese, and Japanese often yields different examples, assumptions, and cultural perspectives. For research and decision-making, multilingual prompting may be an easy way to avoid a single AI-generated worldview.
A new MIT study had 272 experts assess 24 categories of AI risk. Under business-as-usual conditions, 18 of them carry a greater than 10% chance of catastrophic outcomes — a million-plus deaths, $100B+ in losses, or civilization-scale harm — by 2030.
ABBYY is committed to supporting its customers operating in complex regulatory environments through a security-first cloud governance model that aligns with the BSI C5 framework.
ABBYY is the winner of the “IDP Solution Provider of the Year” award in the 9th annual AI Breakthrough Awards program. ABBYY Document AI stands out in the crowded AI market for one simple reason: proven results, with or without the use of LLMs or generative AI.
ABBYY Ascend 2026: Key Highlights
If you missed ABBYY Ascend in Nashville or Brussels, read the highlights by Marlene Wolfgruber. The series focused on helping organizations cut through the noise and understand what it really takes to turn AI initiatives into measurable impact.
Risky Future Demo Day, Hosted by Insurance Journal – July 8 – 1 PM ET
Join ABBYY for this fun, virtual event featuring actionable demos for the insurance industry. The topic for this session is “AI Tools for Underwriting”. It’s a free event open to all insurance organizations.
Webinar: Automate Invoice Processing for Workday with Intelligent Document Processing – July 15 – 11 AM ET
Join Initium SoftWorks and ABBYY for a practical webinar built for Workday users who want to automate invoice processing and other document-driven workflows through out-of-the-box integration with Workday Financial Management.
ABBYY Ascend is heading to Tokyo, Bengaluru and Singapore
As part of the ongoing event series, ABBYY Ascend will bring together enterprise AI leaders who are ready to turn complexity into clarity. Expect new product announcements, expert-led discussions, and unique insights tailored to each market.
Meet ABBYY at Tech for Industry Show – Paris, 23–24 June
Join ABBYY at Tech for Industry Show on June 23–24 in Paris. Visit us at Stand F10 and attend our session, "Documents critiques, décisions critiques : pourquoi l'industrie adopte ABBYY, l'IA documentaire de confiance", on June 23 at 15:20.
You have received this e-mail because you are an ABBYY customer or partner, have requested further information on our products, solutions or services, registered for an ABBYY trial version, and/or agreed to receive news from ABBYY.
