ENTERPRISE

How new context models fix AI's blind spot

Celonis has already helped half of the world's 200 largest companies gain the intelligence to better understand their businesses. Now it's helping them predict the future. 

On Tuesday, Celonis announced the Celonis Context Model, or CCM, which can act as a digital twin of your business operations and dynamically update in real time. The idea is to create a context layer in your organization that is much smarter and more adaptable than just a knowledge graph or an MDM system. 

CCM brings together structured and unstructured data from every interaction across screens, system logs, policies, and process descriptions. This creates a consistent language about your organization that AI can understand. 

Alex Rinke, Co-CEO of Celonis, told The Deep View that the biggest issue with enterprise AI is the lack of context. The best enterprise AI is a “specialist in your business,” rather than a general model. 

"You need to teach it how to be good at doing a specific job in that business,” Rinke said, “That requires a lot of context, and that context needs to be brought together and put in a language that the AI can understand and effectively reason about."

That context can be the missing piece between an AI project that's struggling to create value and one that can automate processes that have been resistant to AI so far. 

Additionally, Celonis announced its acquisition of Ikigai Labs, a company led by Dev Shah and founded at MIT that has spent the past 20 years building AI systems for prediction, forecasting, time-series analysis, and what-if scenario simulation. 

Rinke told The Deep View that Ikigai allows it to bring in graphical models and other tools that specialize in future reasoning, predictions and forecasting. That way, its context model “doesn’t just represent in real time what’s going on right now, but can also look into the future, [at] what could happen.” 

So what can all of this do? In evaluations run by Celonis and customers such as AstraZeneca, Mercedes, Domino Foods, insurance companies, and others, they found that integrating the context layer with AI agents dramatically improved efficiency, accuracy, task completion, and speed.

When it comes to successfully adopting AI in business, smarter LLMs can only take you so far. The context layer can certainly be the difference between an AI project that fails to deliver ROI and one that meets expectations for creating value. Celonis was already doing that with its focus on process intelligence, where it was recently crowned as the industry leader in Gartner's Magic Quadrant. In many ways, CCM and the context layer are simply a rebranding of Process Intelligence. But integrating Ikigai is a game-changer. Where Celonis can make enterprises much smarter about the way they run in real-time, Ikigai can take that data and model out the future so that AI agents can make better decisions when doing things like offering sales, deals, price quotes, and other actions to save time and money and give customers a much faster and better experience, for example. Those are the kinds of outcomes that can compound significantly over time in many organizations, especially the largest ones. 

TOGETHER WITH CHECKSUM

Self-Healing AI Testing Can Save You Thousands. Here’s How:

The team at Clearpoint Strategy knew something had to change. Like a large proportion of engineering departments out there (63% to be exact), they were shipping code faster with AI… but production incidents were increasing, too. And that’s not good.

So they turned to Checksum AI.

This AI-native continuous testing platform auto-generates and self-heals your E2E test suite, runs inside your existing CI/CD pipeline, and keeps pace with the velocity you're already getting from AI coding agents. Translation: It constantly tests your code, so your teams can move fast, stress less, and save money – to the tune of $500K a year for Clearpoint. 

See how it changed everything for them, and try it yourself for free right here.

CONSUMER

Android's new AI trades flashiness for smarts

True to tradition, Google kicked off the one-week countdown to its biggest event of the year, Google I/O, with the Android Show. And this time, AI took center stage in a different way.

Now packaged as Gemini Intelligence on Android, on Tuesday, Google introduced several AI features that will roll out in waves to users, starting with the latest Samsung Galaxy and

Google Pixel phones this summer. All the features are centered on infusing AI in a more proactive and helpful way to tackle everyday issues and save time.

“Gemini Intelligence is essentially a suite of features that brings the best of Gemini on our most advanced devices,” Mindy Brooks, VP of Product for Android, told The Deep View in an exclusive interview. “So with Android fundamentally evolving from an operating system to an intelligence system, your device can truly understand you, work to translate your intention into action.” 

A prime example is Brook’s favorite feature, Rambler, built directly into Gboard to better interpret voice dictation and convert it to text more quickly and accurately. For instance, it will ignore distractions such as “ums”, “likes”, and self-corrections and instead keep the relevant parts to compose the message. It can also seamlessly switch between multiple languages. 

Brooks said, “I use it all the time, and I find it so incredibly helpful. It just cuts so much time off.” 

From its description, it sounds very similar to Wispr Flow, a tool that has gone viral for the ease it gives users for voice dictation, except that Google’s alternative would have a major advantage in native integration. Wispr Flow is great on Mac but isn't as good on Android and iOS due to limited integration. 

The Task Automation feature, which garnered a ton of attention in January, at the launch of the Samsung Galaxy S26 lineup (and also available on the Pixel 10) for actually taking action for you within your apps, is getting an expansion. While the details remain vague, it will apparently be able to do more for you across more apps, as well as take your screen or an image into context.

Other upgrades include: 

• New UI: Gemini Intelligence has an updated design language that builds on Material 3 Expressive 

• Autofill: It can now adapt to fill out more online forms in real-time based on your Personal Intelligence.

• Custom widgets: Build entirely custom widgets using natural language, such as status updates on orders and specific info from your favorite sports teams.

For more about the launch and to hear the full interview with Brooks, you can tune into our podcast episode.

It's refreshing to see Google focus on delivering AI features that actually solve everyday user problems. Google had already been pretty good about this with the launch of the Pixel 10 last summer, becoming what most phone reviewers (including me) felt comfortable calling the best AI-powered smartphone. Brooks echoed that sentiment in our interview: “We don't need to talk about AI; that doesn't need to be the core of it. It needs to be about what are [the] user pain points we [all] have today.” We'll see how all of the features work in action, but the emphasis on things that save time and reduce friction over ones that produce flashy demos is a nice step forward.

TOGETHER WITH VIKTOR

Three departments. One coworker. Same Tuesday.

By 9:14 AM Tuesday, Viktor had already drafted a Q2 board update for the CFO, left line-by-line comments on an engineer’s pull request, and shipped a landing page for the marketing lead’s product launch.

Three different departments. None of them knew the others were messaging the same coworker.

Viktor lives in your Slack and connects to 3,000+ tools including Stripe, HubSpot, Klaviyo, Linear, GitHub, and Shopify. It ships real work like PDFs, PRs, landing pages, and overnight alerts when something breaks.

More than 13,000 teams already run on Viktor. As Joyce Weeks, CEO of Dental Ed, puts it:

“It runs my business behind the scenes.”

Get started free and receive $100 in credits. No credit card required.

CULTURE

A solution emerges for Hollywood's AI problem

With generative AI threatening to upend creativity and causing copyright chaos, the entertainment industry is stuck between a rock and a hard place. 

On Tuesday, several Hollywood stars formed an alliance backing a new standard that gives actors more control over how their likeness is used by AI systems. The protocol, called the “Human Consent Standard,” allows people to set terms for the use of their work or license within AI systems, ranging from restricting use entirely to allowing full, unlimited permissions. 

The standard was launched by RSL Media, a nonprofit co-founded by actress Cate Blanchett, which has received support from a number of high-profile actors, including Viola Davis, Kristen Stewart, Tom Hanks, Meryl Streep and George Clooney. 

The standard is essentially embedded into the underlying digital work, identity or character wherever it appears, requiring web and AI crawlers to check that piece of IP against a registry that’s launching in June. RSL Media then translates those terms into signals that an AI system can understand.

RSL’s standard seeks to quell the chaos surrounding AI’s impact on copyright laws and consent about how a creator’s content is used. 

“AI technologies are expanding rampantly, essentially unchecked and unregulated,” Blanchett said in a statement. “In order for humans to remain in front of these technologies, consent must be the first consideration.” 

The standard also adds to a growing movement in the entertainment industry to create protections for creators against AI. Actors and artists have banded together in the last several months on initiatives including the “Stealing Isn’t Innovation” petition and the Creators Coalition on AI. Meanwhile, major Hollywood institutions are still sorting out what AI means for awards season.

And while the battle against AI ensues, some in the industry are reportedly helping to feed it: On Tuesday, a WIRED report found that Hollywood screenwriters, seeking work, are increasingly turning to AI training to make ends meet. 

AI in the entertainment industry is essentially the ouroboros. A vast proportion of creators are rebelling against it, for fear that it could rob them of their livelihoods. That, in turn, could worsen the AI models themselves, which feed on high-quality art to create their own mimicry and constantly need new data to stay current. On the flip side, if AI does manage to put artists out of work, the models will worsen without authentic training data to learn from. Either way you cut it, it’s a lose-lose situation. While there are responsible ways to use AI in any industry, in entertainment, the people most directly affected are asking that human creativity and control be prioritized. 

LINKS

• Anthropic reportedly in talks to raise $30 billion at $900 billion valuation

• Meta sets annual Connect conference for September 23 and 24

• Google DeepMind demos reimagined computer mouse pointer 

• AI drug firm Isomorphic Labs raises raises $2.1 billion

• Cisco open-sources its specification for agentic security evaluation 

• New York Times issues warning to freelancer writers about AI use 

• Krea 2: The AI creativity firm’s first foundation model, built for “aesthetic diversity and stylistic control.”

• Claude for the Legal Industry: Anthropic announced more than 20 new MCP connectors that link Claude to legal software, and 12 new plugins for specific legal work. 

• SAP Autonomous Enterprise: SAP unveiled a new business AI platform for creating, contextualizing and governing agents.

• Odyssey PROWL: World model firm Odyssey reveals reinforcement learning agents for game environments.

• Google DeepMind: Research Engineer, Frontier Safety Loss of Control

• Visa: Staff Research Scientist

• Microsoft: Research Intern - LLM Acceleration

• Perplexity: Member of Technical Staff (AI Researcher)

GAMES

POLL RESULTS

Have you tried Claude Code or Claude Cowork Yet?

Yes (29%)
No (63%)
Other (8%)