B2B ad examples

Run Opik test suites in CI so every prompt, model or tool change is checked against the same cases before release.

Find the step that failed, turn it into a test case, and run the whole suite on every change before it ships.

It's about time Fin - Intercom's agent was given a face. I was demoing Anam during New York Tech Week at an Intercom event. So decided not just to give Fin a face, but have him join a Zoom call as a participant and let him take the questions himself, live. He explained what he does, and why agents need a face. It's pretty cool to watch. Think about every business that runs on live conversation - sales or support calls, onboarding, coaching, tutoring. Now their users are in a video call with an avatar that interacts with them like a colleague. And it's finally really easy to build an interactive avatar who can join Zoom calls, interact with your users and become the face of your product - you don't need to be a developer!

From asset maintenance to regulatory analysis, not every AI task needs the most expensive AI model. Haystack Enterprise Platform gives you visibility into AI spend and the ability to optimize cost and performance as adoption grows.

+75% of Betsson's incoming iGaming support tickets, resolved end to end by AI. That is after 30%+ of tickets were already solved proactively, before they were even created. On top: 80–90% of KYC issues resolved by AI, and a 3.3 min average resolution. The case study shows what it took in a regulated operation, and who runs it: the support team. Free download for teams at licensed iGaming operators.

New iGaming case study: Betsson now resolves +75% of its incoming support tickets with AI, end to end. Over 30 days, 30%+ of tickets were solved proactively, before they were even created. Of the remaining tickets, +75% were resolved by AI. 80–90% of KYC issues too, 24/7. The case study shows how: one platform with central governance, rules written and owned by the support team, every reply traced to the rule behind it. Free download for teams at licensed iGaming operators.

I've been going through the color explanations from our AI taste study (all 563 of them) and one pattern keeps showing up. A lot of them don't really say what the color is for. They say what it isn't. "Without the coldness of hospital blue." "Rather than corporate-cold." About 36% of them are written that way. Take the telemedicine brief, for example. The models were trying hard not to look like a hospital, and 8 of 10 landed on the same deep teal. Two of them were one hex digit apart. It's a good reminder that "avoid the cliché" is also something a model can learn, and once they all learn it, it stops being an escape. This blog dives deeper: https://lnkd.in/g7UuJdJf

We gave ten AI models the same ten website briefs, from a boutique hotel to a telemedicine app, then asked people to pick the better design, two at a time. 4,500 votes later, here's where every model landed. All 100 designs are live on the page, along with the rankings and the raw data.

You have a limited budget for collecting hand data for physical AI. What do you spend it on? 10,000 more clean demonstrations of pick-and-place and grasping? Or a few hundred hard sequences: fingers self-occluded or hidden behind the object, hands truncated at the frame edge, off-nominal camera viewpoints, rare grasp types? What’s the trade-off in clean vs hard in terms of performance. We would like to put on the table at our research session at Humyn Labs this Saturday. We’re bringing researchers together to examine real failure cases and debate what the next round of hand data should prioritise. 24 October · Indiranagar, Bengaluru Researchers in hand pose estimation, egocentric vision and robot learning, this one is for you! Link to apply in the comments. Sandeep Reddy Sabbella, PhD Ishank Gupta Manish Agarwal Bhoomi Arora Aachman Maheshwari

Theo is an AI workspace for researchers: literature, derivations and code in one place. Sign up free and get 1,500 credits.

Theo is an AI workspace for researchers: literature, derivations and code in one place. Sign up free and get 1,500 credits.

100+ applications already. This Saturday, we're bringing together some very interesting folks: → An Adobe researcher who worked on egocentric hand tracking at CMU → A former CTO who built robots to help astronauts in space → A builder making mocap gloves that can sense touch → A PhD researcher at IISc's HiRO Lab An evening led by Dr. Sandeep, discussing all things egocentric hand tracking and where gaps are. Building in this space? Last few spots. We close applications in 24 hours. Sandeep Reddy Sabbella, PhD Ishank Gupta Manish Agarwal

A smile score isn’t the story. Context is. Hume measures expression across face, voice and language in real time, so your product responds to how people actually come across, not just what they type or say.

“Fine” can be a yes or a goodbye. The transcript can’t tell the difference, and most AI reads only the transcript. Hume measures how it was said: tone, energy, delivery, across voice, face and language, in real time.

Introducing Exa 3.0, a step change in search for agents. Exa 3.0 led on every benchmark we tested, from public research evals to coding and GTM intelligence. The gains come from upgrades across our retrieval stack: our largest embedding model, Dynamic Highlights that use 95% fewer tokens than full pages, and billions of new documents covering products, local businesses, flights, sports, and weather. We are also launching Exa Lite at $4 per 1,000 requests for cost-sensitive apps. For higher quality, Exa Auto spends more compute on every query. Together, they set the new Pareto frontier for search. For coding agents, financial research, and AI assistants answering everyday questions, Exa 3.0 is state of the art.

AI agents spent 23 hours probing major companies. Nobody on the receiving end noticed. That’s the story leading this week’s AgentShock. Microsoft’s new answer is to give agents a live risk score and take away write access when they start acting strangely. I’d like to know who sets that threshold. Also in this issue: Anthropic signs for 420 MW of wind-backed capacity, and an 18-month data center pause in Memphis dies on a 6–6 vote. If an agent spent tonight probing a vendor from your network, which log would tell you? Read the full issue below. 👇

Sovereign AI fails the moment a developer decides the public API is faster than your private GPU cluster. Owning the hardware only protects your data if teams use it. In this post we cover how Klique runs on SUSE AI Factory with NVIDIA so the governed path is also the fastest one. SUSE and NVIDIA provide the Rancher-based infrastructure, models and serving. Klique adds self-service access to approved models and GPUs, quotas that decide which team gets accelerators, an LLM Gateway that logs tokens by user and model, and chargeback so finance can see what each project spends. Learn more about the solution in our latest blog post. At NVIDIA GTC Berlin? If you missed Damian Erangey's demo in the SUSE booth, drop by our booth #7054 and ask for an encore! We still have a few slots left on on our meeting schedule. Blog post and sign-up links are the comments!

Capitol is AI for consulting, and we think it should produce more than answers. It should produce the work behind them: research you can audit, repeat and put in front of a client. Here’s a public example. Five companies will spend the equivalent of 2.5% of US GDP building AI this year, America’s biggest bet since the Louisiana Purchase. So we asked Capitol how America’s big bets have turned out, from the Erie Canal to the Interstates, and published what it found. Some paid off fast. Some took decades. Some overshot. Every number traces to its source. The method is the same for every bet. And it’s designed to be read. That’s what Capitol builds for consulting. See the numbers: https://lnkd.in/eSGwmGf2

AI agents don’t announce their mistakes. Most enterprises can’t see which decisions are safe until one becomes a business risk. The Trusted AI at Scale paper shows how leading organizations score every AI decision, route uncertain ones to human experts and keep a full audit trail from data to deployment.

Planning a major AI deployment means aligning GPU capacity, location and delivery dates well before workloads go live. CambridgeNexus has secured NVIDIA Vera Rubin allocation, with leasing expected in mid-2027. Reservations are open now for whole-rack, single-tenant deployments, with CambridgeNexus operating the infrastructure for you. Explore NVIDIA Vera Rubin reservation options for your 2027 capacity plans.

Many AI teams are looking for greater predictability when planning capacity for training, inference and production workloads. CambridgeNexus has secured NVIDIA Vera Rubin allocation, with leasing expected in mid-2027. Reservations are open now for whole-rack, single-tenant deployments, with CambridgeNexus operating the infrastructure for you. Explore dedicated NVIDIA Vera Rubin rack leasing and how to reserve capacity.

🧑💻 Accounting workflows spread across PDFs, Excel, emails, and specialist systems, with employees repeatedly transferring and re-entering the same information. Sounds familiar? We faced this challenge. We worked with the client to identify where AI could deliver the most value without disrupting existing systems or removing essential human oversight. In just around a month, we mapped 4 accounting processes, identified 10 automation opportunities, and built 2 working AI automations. One of them reduced reconciliation preparation from hours to approximately 20 minutes. But the most interesting part is how we approached automation despite the integration constraints. Read the full case study to explore the implementation decisions, results, and lessons learned: https://lnkd.in/dt_-hgBz/?utm_campaign=%255Btraffic%255DMaincampaign&utm_source=linkedin&utm_medium=paid&utm_id=1206642164&hsa_acc=508184568&hsa_cam=1206642164&hsa_grp=905118884&hsa_ad=1606109564&hsa_net=linkedin&hsa_ver=3 #AIAutomation #CoPilots #WorkflowAutomation #PE

Dubai is turning AI ambition into action, and we're excited to be part of the conversations shaping what's next! C5i is heading to the Dubai AI Festival 2026, with Anees Merchant (Executive Vice President) and Shreyas Sundaresh (Head of AI Transformation & Client Success) representing us at the event. As organizations across the UAE accelerate their AI initiatives, an important question is emerging: What does it really take to make AI work at enterprise scale? At C5i, we bring experience from working with leading global enterprises and financial institutions, helping them navigate complex business challenges through AI, analytics, and intelligent automation. From modernizing enterprise data and transforming banking operations to identifying high-value use cases for agentic AI, we've seen how the right combination of technology, business context, and execution can make a meaningful difference. We're looking forward to exchanging perspectives with the region's business and technology leaders on their AI priorities, the challenges they're navigating, and the opportunities ahead. Hear from Anees Merchant on why the Dubai AI Festival is an exciting opportunity and what we're looking forward to. Whether you're exploring new AI opportunities or looking to scale existing initiatives, connect with Anees and Shreyas. We'd love to meet you in Dubai! Hemant Julka | Parag Penker | Samer Hamadneh #C5iHumanAIImpact #MiddleEastBanking #UAE #BankingAI #ResponsibleAI #DataAndAI #AgenticAI

Another AI pilot won’t fix what’s holding enterprise AI back. The real challenge is making AI work across business processes, enterprise data, and legacy systems, with the governance needed to scale it. That’s the problem C5i is built to solve, combining human expertise, data, and agentic AI to take enterprises from isolated use cases to AI that runs inside core operations. At the Dubai AI Festival, Anees Merchant and Shreyas Sundaresh will be meeting business and technology leaders to discuss what it takes to deliver value beyond the pilot stage. Want to connect? Fill out the form to book a meeting with our team.