RESEARCH

Gartner's AI predictions address the control gap

It will come as a surprise to no one that AI is outgrowing the systems designed to manage it. 

At the Gartner IT Symposium/Xpo this week, the analyst firm offered both its top 10 strategic technology trends for 2027 and its top 10 strategic predictions for 2027 and both lists shared some common themes: AI is making software faster to create and work easier to automate, while making costs, security, and accountability more difficult to govern.

The two lists are distinct, but I've synthesized them into a set of insights to watch for 2027:

• AI's capabilities are outpacing its controls: While AI is letting employees hand off more work to machines, it's also forcing leaders to rethink how they control costs, grant access, and decide who's responsible when things go wrong. 

• More autonomy needs better accountability: The systems for managing AI need to catch up with its growing autonomy. By 2028, 25% of enterprise security breaches will trace back to AI agents with mismanaged credentials, Gartner predicts. Honestly, that number seems conservative. 

• Disposable software will change the enterprise: AI makes software far easier to create. Gartner predicts that by 2029, 80% of all new applications will be used for less than a year and made to be intentionally disposable. Still, disposable applications don't make the underlying data, integrations, or security disposable.

• Better context becomes a business advantage: Even the most powerful model still needs to understand how your organization works to be effective. Organizing your enterprise knowledge so AI can use it will become as consequential as choosing the right model. By 2027, companies that prioritize making their data AI-ready will increase agentic accuracy by up to 80% and reduce costs by up to 60%, Gartner predicts. 

• AI spending becomes a security issue: Gartner warned that lower price per token does not guarantee a better business outcome. Companies need to focus on the cost to finish useful work, including the retries and mistakes the models make along the way. In fact, Gartner warned that attackers could use this against companies, warning against "cost exhaustion" attacks that drive excessive AI consumption to run up your bill. Gartner predicts that 80% of companies with public-facing AI will experience such an attack by 2030. 

Nevertheless, Gartner analyst Daryl Plummer in his presentation on the 2027 predictions at the event, warned, "Your worst action is when you focus strictly on cost reduction. Don't do that. Cost reduction is the denominator. Value is the numerator. You have to worry about both of them to make the equation go the direction that you want."

Plummer also dealt with the elephant in the room: the impact of AI on jobs. He told the crowd full of executives and directors, "Don't think 'replace workers.' Think 'augment and extend.' That's the better way to go."

The organizations that will benefit the most from AI will be the ones that can clearly explain what AI is doing and why, and exactly how the outcomes justify the costs. While that should sound obvious, there's been so much emphasis on AI adoption for the past few years that many enterprises have kicked the can down the road on ROI. Gartner believes that period is over, and organizations have to get more disciplined and more systematic about AI transparency, accountability, and governance. As I mentioned in our report on Gartner's surveys of executives on AI, only 11% of CFOs said they are seeing clear returns from AI and only 13% of CIOs say their organizations have seen transformational value from AI. These numbers are going to have to rise for stakeholders to continue to believe in AI and keep funding investment. For its part, Gartner's predictions offered actionable insights and mental models for enterprises to adopt to get this moving in a better direction.

TOGETHER WITH AMD

AMD Helion™ Agentic AI Rack

AMD Helion™ Agentic AI Rack is an open, rack-scale reference design purpose-built for agentic AI and powered by AMD EPYC™ 9006 Series Server CPUs.

Configurations built for your agentic needs, from a CPU-only rack designed for hyperscale deployments to an integrated CPU and GPU rack for enterprise environments, both built on OCP standards.

Discover how AMD Helion brings choice and flexibility to your agentic AI needs.

PRODUCTS

OpenAI's Jev rival bets on simplicity

OpenAI may be reading the tea leaves on enterprises' pivot to efficiency. 

On Thursday, the AI lab made Decisions API, its competitor to Typesafe AI's recently popular Jev, generally available to users for the first time. The company is offering Decisions API at 10 cents per million input tokens, one cent per million cached input tokens, and not charging for output tokens—a fraction of what it charges for its frontier large language models. But that's still over twice the cost of Jev.

The company laid out a number of ways it's eating its own dog food, including the team behind its recently released Dots agents using Decisions API to reduce response times and improve interactivity. 

OpenAI's Decisions API is functionally quite similar to Jev, creating output that software can act on, rather than focusing on human-readable outputs. The difference is in the models that lie beneath the surface.

• Decisions API is built on top of GPT-6 Luna, the company's most cost-efficient and lightweight model, and is framed by OpenAI as a specialized API. Though the company may work on specific training for these kinds of models, one OpenAI product lead told TechCrunch that it's "just plain old Luna that gets the job done." 

• TypeSafe, meanwhile, developed an entirely new class of model for Jev that isn't a language model at all, instead using a tactic called  "Reinforcement Learning for Calibrated Decisions," which answers queries with "epistemically honest probabilities," rather than "programmatically verified" outputs. 

However, both releases target the same use case: rapid decision-making for software to replace generative responses. And both have seen success in saving time and cutting costs. OpenAI said that, internally, their offering has allowed agents to make near-real time decisions up to 10 times faster than GPT-6 Luna on its own through the Responses API. TechCrunch reported that it cut decision making time from three seconds to as low as 100 milliseconds. 

Comparatively, TypeSafe reports that Jev is two orders of magnitude faster than existing models, with an end-to-end response time between 70 and 500 milliseconds, compared to up to 329 seconds for existing LLMs.

TypeSafe's Jev has skyrocketed in recent months as enterprises pivot from shelling out on expensive frontier large language models to more efficient alternatives. Investors, too, are taking an interest: Just weeks after its initial launch, the startup raised $870 million at a $7.5 billion valuation in a round led by Andreessen Horowitz. In their funding announcement, the company said that a third of Fortune 500 companies are "getting their Jev on," and that its product has already saved customers millions of dollars in production costs.

It only makes sense that OpenAI is leaning into efficiency at a time when enterprises are tightening their purse strings. And seeing the success of Jev (from a former OpenAI researcher and ChatGPT inventor, no less), it's natural for the lab to use the same fundamental tactic of limiting generative responses in favor of software-focused outputs, especially as agents eat up a growing number of tokens. Plus, the fact that all OpenAI had to do to achieve a similar effect was modify its existing Luna model is certainly an added bonus. However, despite Jev's rapid rise to fame, OpenAI may have a few advantages over the startup. For one, the name recognition and market share that OpenAI already has with enterprises could make it easier for companies to trust the tech. Additionally, because OpenAI's models are already integrated within a number of enterprise workflows, it's likely that the Decisions API will be more seamless to add to the stack.

TOGETHER WITH CREATIO

Enterprise AI has stopped experimenting

A year ago, most leaders said AI agents would become critical “soon.” Soon has arrived.

Creatio surveyed 600+ IT and business decision-makers for its second annual State of Enterprise AI Adoption Report, and the shift is striking:

• 70% of organizations are already past the experimentation stage

• More than three-quarters plan to scale AI agents across departments in the next 12 months

• Only 17% qualify as true AI Adoption Leaders

What sets those leaders apart? They automate more workflows, scale agents across teams, and build strong governance. In fact, organizations with established AI governance are more than 4X as likely to feel very confident scaling AI. 

And the workforce story isn’t what you’d expect: just 9% anticipate significant headcount cuts, while 89% expect AI agents to expand capacity, augment teams, or create new roles.

See how your organization stacks up.

Download the report

RESEARCH

How to turn everyday frustrations into AI wins

Your next big AI win may come from someone your leadership team would never think to pick.

In this episode of The Deep View Conversations, we sit down with Todd Earwood, CEO of Betra Works and an AI investor at HalfCourt Ventures, to explore what it takes to deploy AI for useful work inside organizations.

Drawing on his work with 1,200 leaders and executives, Todd explains why the people closest to the work often see AI opportunities that executives miss. He shares how a university athletics sales rep used AI to uncover new ticket buyers, how a warehouse employee helped a manufacturer rethink a 22-year-old spreadsheet, and why a skeptical, experienced employee became an unexpected AI champion.

We also discuss Todd's book, Adapt & Adopt, and his approach to teaching people how to identify useful AI tasks, evaluate opportunities, and build momentum from individual wins to broader adoption.

Other topics covered:

• Three paths to putting AI to work in a business
• Why employee frustrations can reveal valuable AI opportunities
• The gap between personal AI use and workplace adoption
• What Todd learned from investing in more than 100 AI companies
• Building a tech career outside San Francisco and New York
• Why meeting recorders and voice dictation can make an immediate difference
• Using Claude to protect your calendar and priorities

If you're leading a team, trying to make AI useful in your own job, or deciding where to start, this conversation offers practical ways to connect AI's capabilities with the work people do every day.

📺 Watch on YouTube
🎧 Listen in your favorite podcast player

Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology: https://tdv.transistor.fm/

LINKS

• OpenAI lays out protocols for securing internal models in training

• Anthropic, Hg strike deal to launch European consultancy

• Biotech startup Pathos AI raises $237 million, $3 billion valuation

• US Treasury proposes crack down on tech investments in China, Russia

• OpenAI reportedly eyes $30 billion round at $1.25 trillion valuation

• AI coding firm Replit reportedly shifts AI spent from Anthropic to OpenAI

• Kimi K3.1 Pro: Moonshot's most capable model yet, with 1 million token context, 2.8 trillion total and 104 billion active parameters. 

• Canva Craft: The design platform's first image generation model, designed to compete with image models from frontier labs. 

• Carbon-X: From Hugging Face, an open 14-billion-parameter model for DNA understanding and generation.

• Seedance 2.5 4K: The most nuanced Seedance model yet, with improved facial expressions and movements and richer visual detail.

• MiniMax M3.1 Flash: Open weights now publicly available, with 428 billion parameters total, 23 billion activated per token.

• Rowspace: Applied AI Engineer

• Coframe: Agent Engineer

• Photon: Artificial Intelligence Engineer

• Samba: AI Product Engineer

GAMES

A QUICK POLL BEFORE YOU GO