Every week, CX Network publishes a round-up of funding news, product launches and exits from start-ups with solutions for CX, customer service and workplace challenges.
With details of the who, what, how, when and why, this new bulletin gives CX leaders and practitioners an overview of the latest capabilities in the vendor market, who the biggest new players are and how they can solve long-standing challenges with new solutions.
This week's bulletin covers:
- Guava's $12.5 million Seed round aims to tackle voice AI's growing pains
- Phonic's $27.5 million Series A round to help voice agents with understanding
- Firmly's $15 million Series A to give retailers control of agentic commerce
- Larridin's $30 million Series A to connect AI spending to employee productivity
- And Listen Labs' agentic twin launch following its acquisition by Salesforce
LA-based Guava raises $12.5m Seed to tackle voice AI's "real-world growing pains"
Guava, a new developer platform for building voice agents, this week confirmed it raised US$12.5 million in Seed funding, with the round led by Acrew Capital.
Participation came from Flourish Ventures, WillowTree Ventures, Clocktower Ventures and Guava's pre-seed investor, Commerce Ventures.
Guava was founded to solve a specific problem with voice AI: most voice agents are stitched together from components provided by different vendors and controlled largely through prompts.
Instead, on the model side Guava builds and runs its own speech recognition, language models, and speech synthesis as a single stack.
On the control side, the SDK is designed to help developers integrate voice directly into their apps via business rules and code, rather than forcing them to become experts in managing voice agents and prompts.
Developers can sign up for a free version and have a working agent in under five minutes, while enterprise customers typically take their first live regulated call within 14 days of kickoff.
Guava's first public model, Daytona, launched in September and is already supporting more than a dozen enterprise customers, powering tens of thousands of calls per day.
Co-founder and CEO Omri Dahan, said: "Nearly every company we spoke with was dissatisfied with the voice AI they'd tried. It sounds fine in a demo but breaks down on real, high-stakes calls. We built Guava to give enterprises agents they can trust with high-volume, high-stakes conversations. Now we're building the go-to-market team to take Guava to every enterprise and BPO ready to move voice AI from pilot to production."
The fresh capital will support continued model and product development and kick-start Guava's go-to-market team serving enterprises and business process outsourcers.
Phonic raises $27.5m to help voice agents better understand callers
The other problem with voice AI is its ability to understand what the customer on the other end of the phone is saying.
Phonic, a two-year-old San Francisco startup that builds speech-to-speech AI models for voice agents to understand callers, raised $27.5 million in its Series A round. The raise was led by First Round, with Lux Capital, Felicis, Definition, and Essence VC participating. It's also backed by angel investors including Erik Bernhardsson (Modal), Clement Delangue (Hugging Face), Arash Ferdowsi (Dropbox), Amjad Masad (Replit), Gokul Rajaram, Guillermo Rauch (Vercel), Naval Ravikant (AngelList), and Qasar Younis (Applied Intuition).
Launched in April 2025, Phonic's platform builds on a similar problem to that outlined by Guava: most voice agents today are assembled from parts. One model transcribes the caller, another decides how to respond, and a third speaks the response aloud.
This led Phonic to double down on three key capabilities for its solution:
- Phonic-1 was post-trained for natural conversation, so it understands tone, handles turn-taking in noisy environments, and responds the way a person would.
- Other models wait for the caller to finish, say "let me look that up," and only then call a tool. Phonic-1 is the only model that can act while the caller is still talking, so the answer is ready when they stop.
- So, it's fast. But in addition to working while the caller speaks, in end-to-end full-duplex benchmarks, its developers say it "stands alone on the Pareto frontier of intelligence and latency".
It's also multilingual.
Calling 2026 its "breakout year", Phonic's founder said in a press statement that whether users are in customer support, healthcare, recruiting, and logistics, "teams that have switched to Phonic are seeing higher customer satisfaction and the ability to handle more complex tasks".
On October 29, the start-up unveiled Phonic-1, the "most advanced speech-to-speech model for agentic tasks". Phonic-1 holds the number one spot on the Artificial Analysis Speech to Speech Index and delivers the highest task success rate on agentic tasks at the lowest latency.
Firmly raises $15m in Series A from J.P. Morgan Payments, Mastercard and Samsung NEXT
Firmly yesterday announced a $15 million Series A led by Squarepoint Capital and supported by J.P. Morgan Payments, Mastercard and Samsung NEXT as it looks to solve the biggest pain point of agentic commerce: the retailer's lack of control.
It also unveiled Ask Agent, an end-to-end conversational shopping experience for merchant websites that includes AI-powered natural language product discovery with a real-time catalog, in-chat checkout and traditional customer support functionality.
Capitalizing on the fast-growing AI search traffic landing on retail websites – which Adobe predicts will increase 130 percent this golden quarter – Firmly provides a common infrastructure layer enabled by its Connect platform, which creates a single no-code connection.
This allows merchants to sell through new channels and maintain control of customer relationships, transaction data and their role as merchant of record.
The Series A funds will help Firmly accelerate global growth of its Connect platform with merchants and across AI agents, marketplaces, social networks, shoppable advertising platforms and other commerce destinations.
Todayfirmly Connect can help AI assistants, publishers, social media platforms and other commerce channels access more than 11 million merchants representing more than $900 billion in GMV.
Firmly co-founder and president Damien Balsan, said: "Whether a shopping experience is human-led or agent-led, merchants need one reliable connection to all of it. Today's protocol standards are fragmented and can be slow to implement for merchants — firmly delivers the infrastructure layer that bridges the gap, combining the flexibility of browser automation use by Muse or Instinct with the security, Trust and efficiency of deterministic protocols like UCP, all without a single line of merchant code."
Larridin raises $30m in mission to connect AI spending to employee productivity
South San Francisco based Larridin raised $30 million in its Series A, with the round led by Pelion Ventures and supported by GFT, W23, Thayer, DG Daiwa Ventures, Cougar Capital, and Alumni Ventures, as well as previous investors Andreessen Horowitz and Gradient Ventures.
Helping organizations measure the ROI of AI, Larridin measures how companies' AI spending affects employee and agent productivity. It can deliver insights on what AI costs, what teams produce with it, and which workflows to automate next.
To date, Larridin has measured more than two billion AI sessions and tracked more than 82,000 workflows end-to-end. Across its customer base, Larridin reports a 47 percent average increase in measurable team output and a median $2.4 million AI spend optimized within the first year.
With its latest funding, Larridin will accelerate product development and expand its teams to further build out the capabilities of the industry's most complete AI intelligence platform.
"Enterprise AI has moved quickly from innovation budget to financial accountability, and our investors share our conviction that companies must be able to prove tangible business value from skyrocketing AI spend," said CEO Russ Fradin. "Our Series A funding reflects their deep belief in our technology to solve this critical challenge, enabling us to deliver even greater data insights into every AI-powered workflow across the enterprise."
The new funds bring Larridin's total raised to $47 million.
Listen Labs launches Listen Twins following Salesforce acquisition
Listen Labs has launched Listen Twins, one-to-one AI models of real people in a company's target audience. The news only weeks after Salesforce announced a definitive agreement to acquire the start-up.
Each twin is built from an in-depth interview and can predict what that person will do and why. Companies can ask their own panel of more than 100 twins a question and see how their audience would respond in seconds. By way of example, an organization could ask Which campaign concept should we run? How will people react to our announcement? What would customers pay more for? Or should we enter this market?
"Our Twins know the story behind every answer," said co-founder and CEO Alfred Wahlforss. "Across millions of interviews, we have identified keystone questions that unlock how people actually behave. The ability to predict how people will react is a powerful tool. So, we're building this responsibly."
The capabilities are similar to the agentic twins developed by CVS Health and its partner Simile, or the synthetic data API developed by RYA.
Listen Labs closed its Seed and Series A rounds in April 2025 and, according to reports, was in the advanced stages of running its Series C round, having signed a term sheet for $125 million and securing confirmation that Menlo Ventures would lead the round. If it hadn't been acquired by Salesforce, that round would have valued the company at $1.5 billion. The terms of the acquisition have not been disclosed.