The 14 next big things in applied AI for 2025

Let’s face it: The fact that AI is amazing is no longer all that … amazing. The technology is under ever-increasing pressure to prove its real-world value for consumers, businesses, and researchers in specific contexts. These honorees in the applied AI category are proving AI’s worth for fashion advice, pharmaceutical advice, coding, and much more.

Alta 
For bringing AI to personal styling
For people who lack style expertise or time for outfit planning, the task of  choosing what to wear can be a daily frustration. Alta built a personal AI stylist app that generates outfits based on users’ actual wardrobes, lifestyle, budget, weather, and upcoming events—whether they’re dressing for a board meeting in Switzerland or a summer wedding in Napa. Users can upload their wardrobe or let the app automatically scrape their fashion buys from receipts and photos, then receive personalized suggestions they can visualize on customized avatars. The style agent learns each user’s unique style preferences, getting smarter with use. Alta raised $11 million from top-tier investors including Menlo Ventures, secured a partnership with the Council of Fashion Designers of America that gives CFDA members access to its AI platform, and partnered with tidiness guru Marie Kondo to offer premium closet organization services.

Ambience Healthcare
For freeing doctors from documentation drudgery
Caregivers spend countless hours every week filling out patient charts—time that could be spent on actual patient care. Ambience Healthcare has developed an AI platform that listens in on patient–physician conversations in the exam room via a phone app and automatically generates comprehensive medical notes. The AI then creates a draft summary of the notes—complete with suggested ICD-10 and CPT codes—that the caregiver can review, edit, and sign. The system integrates with major electronic health record systems such as Epic and Oracle Cerner. Cleveland Clinic is now implementing Ambient’s solution after doing a comprehensive head-to-head test of “AI scribe” solutions for healthcare, testing five leading solutions with hundreds of clinicians and across more than 80 medical specialties over six months. 

Bolt
For bringing vibe coding to the browser
Building web and mobile applications traditionally requires multiple development teams, technical infrastructure, and months of coding. Among the highest-profile entrants in the new category of vibe coding tools, Bolt changes app creation by letting users describe what they want to build in natural language then instantly generating code. The platform handles front-end and back-end development logistics as well as hosting without complex setups or cloud services. After launching with a single tweet in October 2024, Bolt’s business grew quickly, scaling from zero to $40 million annualized recurring revenue in just four months. The company secured $83.5 million in Series B funding at a $700 million valuation.

Cradle
For accelerating protein engineering with AI
Traditional protein engineering is slow, expensive, and unpredictable, hindering progress in pharmaceuticals, materials science, and biotechnology. Cradle harnesses generative AI to accelerate protein design by creating entirely new protein sequences tailored for specific functions. This distinguishes Cradle from DeepMind’s AlphaFold, which predicts the structures of proteins. The platform reduces experimental iterations, improves success rates, and uncovers novel structures that were previously out of reach for developing therapeutics, sustainable materials, and industrial enzymes. In 2025, Cradle expanded to 21 customers, including Johnson & Johnson and Novo Nordisk, demonstrating real-world validation of its technology in high-stakes drug development.

GitHub
For making AI coding collaborative
In the age of AI coding assistants, developers are under pressure to ship code faster while maintaining quality, often forcing a choice between speed and control. Since becoming the first widely used coding assistant, GitHub Copilot has evolved beyond simple code suggestions into a tool for creating entire new software features and functions. But GitHub and its parent company, Microsoft, have taken a distinct approach to developing Copilot: Rather than pursuing full automation, GitHub designed Copilot to leave the human coder firmly in control. The assistant works like a good human teammate, GitHub says, showing its work  and asking for review before anything ships. GitHub says Copilot’s user base quadrupled year over year in 2025 and now includes 15 million developers and more than 77,000 organizations. 

Google 
For applying secure open models to healthcare
Healthcare AI developers often struggle to build medical applications because they can’t access specialized models that handle sensitive data securely. Google’s MedGemma family offers the first open multi-modal models trained specifically for medical text and image comprehension, enabling developers to keep sensitive data within private environments while adapting models for specific use cases. The models range from 4 billion to 27 billion parameters, small enough to fine-tune and serve on a single GPU, with the 27B version featuring clinical reasoning capabilities useful for patient triage and differential diagnosis. MedGemma  achieved over 150,000 downloads from more than 10,000 developers in its first month, with the community creating and sharing more than 100 fine-tuned versions on Hugging Face.

Hebbia
For extracting insights from financial document chaos
While making critical risk assessments and investment decisions, financial services professionals often must rely on unstructured data contained within numerous documents. Hebbia custom-built a generative AI solution for finance that lets users more quickly capture insights from millions of diverse documents located all around the organization. An investment banker might use Hebbia to generate conclusions from a call transcript, or someone in private equity could use it to identify potential risk factors before signing an investment deal. Hebbia can also summarize external documents such as market reports and credit agreements. The platform’s versatility across multiple financial tasks helped Hebbia secure a $130 million funding round led by Andreessen Horowitz. Hebbia says it now serves 30% of all U.S. asset managers and has processed over 100 million documents, or 10 times more than its nearest competitors.

Jigsaw 
For making sense of massive public conversations
Governments and other organizations struggle to synthesize thousands of public comments on civic issues. Traditional analysis methods can take months to deliver results, leaving feedback loops open too long for meaningful action. Jigsaw, an incubator inside Google, used Google’s Gemini AI model to create a toolkit called Sensemaker. It identifies key topics and themes from large-scale online conversations, allowing users to understand thousands of perspectives within minutes while preserving discussion richness. The technology surfaces patterns, areas of agreement and disagreement, and actionable insights from data that was previously impossible to process at scale. In its pilot with Bowling Green, Kentucky’s BG2050 initiative—a project addressing the expected doubling of the city’s population by 2050—local leaders used Sensemaker to analyze a four-week online conversation, enabling them to draw insights from community input that would have otherwise remained buried in unstructured data.

Pando
For taming logistics chaos 
Global logistics has faced relentless disruption in recent years, from the Russia–Ukraine war to sourcing upheavals caused by the The President administration’s tariffs. Pando recently added an AI agent called Pi to its platform for managing complexity and risk. The agent is powered by proprietary logistics language models and can automate operations such as requesting shipment from a carrier, invoice validation, and anomaly detection. The agent recommends actions, explains its logic, and executes tasks after confirmation. Within weeks of Pi’s launch, a number of Fortune 50 brands, including Meta, onboarded Pando, demonstrating the platform’s ability to handle the complexities of enterprise-scale logistics.

Samsung Electronics America
For making AI feel natural on mobile devices
Smartphone AI features can sometimes feel gimmicky or disconnected from real-life workflows, creating barriers and distractions rather than enhancing the user experience. The features not only have to be useful but they have to show up at the right times and right places in the UX. Samsung did it right with Galaxy AI, which it integrated directly into its smartphones’ Android operating system to provide context-aware, personalized experiences through multimodal AI agents that can interpret text, speech, images, and videos. Users can perform multistep actions across apps, using plain language to get directions, send messages, and update calendars simultaneously. The Galaxy S25 series implementation has earned widespread praise from tech reviewers—Tom’s Guide remarked, “The S25 Ultra is packed with smarter AI features I wish the iPhone 16 Pro Max had.” 

Sonar
For ensuring AI-generated code meets enterprise standards
AI coding assistants have proved that they can accelerate software development, but as the tools have evolved to touch more and more parts of an organization’s code, they also can introduce hard-to-detect flaws that show up as bugs later on. Sonar’s platform uses AI to scan software for quality problems then fixes them. Fixes are informed by the platform’s deep experience—it analyzes more than 300 billion lines of code every day, the company says. The platform’s AI Code Assurance mode provides stricter quality gates for AI-generated code, while AI CodeFix generates contextual repair suggestions based on precise analysis findings rather than generic recommendations. Sonar says that 70% of developers rate its fix quality at 4 or 5 out of 5.

Typeface
For generating end-to-end marketing campaigns
Marketers too often must choose between AI-generated content that’s generic or that is off-brand. Typeface created the first AI marketing platform that orchestrates the entire content process from brief to finished campaign. While the Typeface integrates with more than 30 AI models, the company trains custom models that maintain brand voice, tone, and visual identity. The platform’s Brand Hub is a searchable AI content repository that enforces compliance and governance guidelines. Spaces provides a visual workspace where marketers create personalized emails, ads, web pages, and videos without becoming prompt engineers. Typeface secured major enterprise deals with Fortune 100 companies, including Asics, in 2025.

Vermillio
For arming creators against deepfakes
Vermillio makes a new kind of IP protection platform that identifies the unauthorized use of a person’s likeness or voice in synthetic or AI-generated audio and video. Traditional content protection systems can fail to detect AI-generated content derived from a face or voice because they’re better at detecting exact replicas of the original content. Vermillio’s TraceID technology assigns digital signatures to every fragment of intellectual property, creating “soft bindings” through digital hashes and fingerprints that aren’t easily removed by generative AI models. Vermillio’s platform tracks IP usage across images, text, audio, and video, ensuring proper attribution and compensation while detecting harmful deepfakes. Vermillio says it had more than 130,000 pieces of unauthorized AI-generated content taken down in Q4 2024 alone. In March 2025 the company closed a $16 million Series A round led by Sony Music Entertainment and including Disney and Warner Music Group. 

Warp
For reimagining the terminal for the AI age
Developers have long preferred to control their machines via a command-line terminal because it’s more direct and precise. Unfortunately, the terminal, which was born in the 1970s, hasn’t kept up with new developments in AI-powered coding assistants and agents. So Warp built a modern, AI‐powered terminal for developers called an Agentic Development Environment (ADE). The environment maintains the command line’s power while adding support for natural-language-based AI coding assistance and the ability to manage multiple AI agents. The result is a platform where AI agents have more visibility into the code base so that they can detect potential problems and offer ways of fixing them. Warp says it has seen 90% year-over-year growth in its user base, with 600,000 active developers now using the platform, including 16,000 engineering teams. 

The companies and individuals behind these technologies are among the honorees in Fast Company’s Next Big Things in Tech awards for 2025. Read more about the winners across all categories and the methodology behind the selection process.

View the full article