“Everything you need is in the ticket.” Famous last words. The history is elsewhere. The last decision is buried in a thread. The detail that makes it all make sense is hiding in a doc. Computer, by DevRev helps connect the context behind the request, so you can figure out what happened before making your next move. Happy Halloween 🎃

You call about yesterday’s failed delivery. The voice agent apologizes beautifully, asks for details the company already has, then transfers you to someone who asks you to start over. That’s a lot of talking for someone who just wanted to know where their order went and it gets annoying fast. No kidding. By then, you’re doing the remembering for the machine. @nejjynejjy , our Voice AI Lead, puts the responsibility where it belongs in her new article for Engage Customer. “The person should not have to become better at talking to the machine. The machine should become better at understanding the person.” Read Jen’s full piece to know more: bit.ly/4iI0HMZ

At DevRev’s leadership retreat, Patrick Duffy, our Head of Creative, presented “Work softer,” the line behind his first brand campaign. It was meant to make people stop and look twice. The first question felt like a win. Then came another. And another. Soon, the room was asking whether “Work softer” needed one definition or could mean something different to everyone. Pat had their attention. But could he bring the room along with the idea? Watch the full episode now: dvrv.ai/48kbHKt

Could you tell Codex what to test instead of writing another fine-tuning script? Ritesh Goru and @ahmedmbashir are exploring that with Baseten models on DevRev Live. We’ll look at the experiment and what a researcher still needs to check. Oct 29, 8:30 PM IST. Join us here on X!

The issue says “fix it.” The repo has questions. Tomorrow on DevRev Live, Ahmed Bashir, Piyush Malu, and Abhinav Manchanda show how we bring customer context into Computer’s coding workflow, from issue to reviewable code, and where a developer steps in. Join us here on X!

Episode 5 of The Computer Program, Season 2 is out! One of the key features planned for Computer’s desktop launch, Multiplayer, had been taken out of scope. It was meant to let teammates continue a Computer conversation where someone else left off, but a blocker and a tight deadline stood in the way. Hours after agreeing on how the experience should work, the team ran into that very gap themselves. They started tackling the blocker, got the go-ahead, and brought Multiplayer back into scope. The deadline hadn’t moved. Could they pull it off? Watch the full episode to know more: bit.ly/4yuJkUJ

An escalation closes, but the work behind the resolution rarely makes it into the ticket. It lives across old conversations, in the memory of someone who remembers the workaround, or in a product change from months ago that finally explains the customer’s issue. Phil Verghis, Founder of the Verghis Group and Advisor at DevRev, calls this the “second job” – the work your best people do just to get their actual job done. When problems cross vendors, that job often falls on the customer. It’s 2 AM, their systems are down, and every vendor’s dashboard is green. AI can connect the old case, the Slack thread, and the engineering change. But the call nobody documented? The workaround only one colleague knows still holds? So how do we make that knowledge available when it matters most? Phil Verghis explores the challenge and what it takes to solve it in the full article: bit.ly/4Bh1LPe

Your AI misses a customer affected by a bug. Would a better model catch it, or did the system never connect the bug to that customer’s account? Our Co-founder and CEO @dheeraj shares what we learned testing that question in this CIO article: dvrv.ai/3TAKSy8

Need a status, comparison, or trend? You shouldn’t have to dig through a long AI answer. Later today on DevRev Live, Technical Engineer Shivam Kumar and Product Designer Kavinas Sundaramurthy discuss how they’re designing GenUI in Computer. See you here when Shivam and Kavinas go live.

“Work softer.” Sounds good. But are we really going to say that? Even inside DevRev, it took a little getting used to. What does it mean? Is it too soft? What happened to working hard? In S2E4 of The Computer Program, the team shares those first reactions... and how a few days with the idea helped them see where it could go. Catch the full story in S2E4: dvrv.ai/4zlHC8I What was your first reaction to “Work softer”?

“I’ve always been drawn to companies that challenge conventional thinking.” We’re glad that brought you to DevRev, Chris. @chriskaddaras is joining us as an advisor, bringing experience from Juniper Networks, @nutanix , and beyond. We’re looking forward to his perspective on the decisions ahead: where to focus, how to grow, and how to turn our ambition into something that matters to customers.

“Got a minute?” adds up across a company. Someone needs access, an answer, a routine fix. They’re stuck until another person can stop what they’re doing and help. With self-service, they can get those everyday needs sorted and get back to work. Your team has fewer interruptions and more time for problems that need their experience and judgment. Our co-founder and CEO @dheeraj shares his take on getting that balance right by knowing what AI can handle on its own and where people need to step in. See how we’re putting that balance into practice with customer and employee self-service at DevRev: dvrv.ai/478HjCh

It’s live! The Computer Program, season 2, episode 3 🖥️ Two months until the brand campaign for Computer, by DevRev goes live. The website copy is drafted. The billboards? Still to be sorted out. The film shoot? Not booked yet. And the team is still asking... what will make someone stop and pay attention? The copy. The billboards. The film. All in two months. Can they make it land before the clock runs out? Watch the full episode here to know more: dvrv.ai/4jaasUR

Your to-do list could use some backup. Preferably one that does the work. AI workers are designed to use business context to plan tasks, make decisions, and take action. So what can you hand over, and what still needs your attention? Join @ahmedmbashir and Nikhil Talwar tomorrow, live right here on X, to see what you can hand over to an AI worker. Bring your questions – and that task you keep putting off!

You bought AI to give your team time back, but how much are they spending getting it ready to help? Before a QBR, your team still has to gather account updates, support tickets, and customer commitments. By the time AI has the context, they’ve already done much of the work you bought it to handle. But what if your AI already had that context when your team asked? That’s where Computer, by DevRev comes in. Its Shared Memory connects your live business data, so it can reason across the full context and deliver precise, cited answers, with permissions intact. See how Computer puts your business context to work: dvrv.ai/478bppw

Mollie Javerbaum from @OpenAI joins our CTO @ahmedmbashir on DevRev Live today! 🎉 What does it take to turn frontier intelligence into AI that gets real enterprise work done? Watch live right here on X in a few more hours!

Ten people loved the pilot. Now a thousand want access. Sales needs the latest pricing. Support needs a customer’s history. A new hire needs the right policy. Each answer means searching through more of your company’s information. And when the answer misses something, someone asks again, adds more context, or checks it manually. You’re paying for another attempt while your team is still waiting. Soon, you’re deciding who gets access and how much they can use. An awkward follow-up to “we’re rolling this out to everyone.” Before expanding the pilot, ask what it costs to get an answer someone can actually use – across every team, every day. See what Computer can do: dvrv.ai/4xUEWOo

It’s live: “The Computer Program”, season 2, episode 2. Three months to launch Computer, by DevRev – to everyone. A brand campaign is taking shape alongside it. And the team is still working out what happens when someone opens Computer for the first time. – Will they trust it enough to connect their data? – Can they set it up without someone beside them? – And when they ask their first question… Will the answer actually help? Martin Jackson, DevRev’s copywriter, has spent months shaping how Computer responds. One of the instructions he wrote says it plainly: “Get to the point. No bullshit.” Now product, design, engineering, and marketing have to make the rest of that first experience feel just as clear – before launch. Can the team make someone’s first experience with Computer as clear as Martin’s words? Watch the episode now: dvrv.ai/47r5jR4

We’re excited to share that DevRev is a launch partner in @OpenAI Marketplace. Getting help at work shouldn’t mean repeating your story, digging through tabs, or figuring out who owns what. DevRev helps customers and employees get unstuck – with the context needed to move things forward. And when the next step needs a person, the right context comes with it. Now, OpenAI enterprise customers can find DevRev in the OpenAI Marketplace, see what it can do, and tell us they want to learn more. Want to know how this will work? Adnan Bhutta, our CVP of Business and Alliances, shares more in his blog: dvrv.ai/4ddOg8C

AI should help people do more without asking them to compromise on trust or security. That’s a belief we share with @Prof_Malhotra ( Professor at Harvard Business School, Founder of Kilmer AI), who’s joining DevRev as an advisor. In his words: “We shouldn't have to choose between AI companies that shoot for the stars and those committed to doing things the right way.” Welcome to DevRev, Deepak. We’re glad to have your perspective as we build.

An AI model can top a leaderboard and still be the wrong choice for a business. Perhaps the work needs to stay in a particular region. Perhaps the cost changes once retries and human review are counted. Or perhaps the next model is better for the task, but switching would mean rebuilding everything around it. So which model wins? It depends on the work. Sometimes a frontier model earns its premium. Other times, an open-weight model meets the bar and gives a company more control over where it runs. The bigger question is whether a company can change models without losing the context and controls it has built. Michael Machado, our CVP for Product & Brand, writes about why that choice matters now: dvrv.ai/4xFVF89

🔇 Going live today: When should AI workers think fast? Join Shanay Mehta and Ahmed Bashir for today's DevRev Live on X as they discuss Jev, TypeSafe AI’s System 1 model, and why fast, bounded decisions belong alongside deeper reasoning.

We’re back with Season 2 of “The Computer Program” – the world’s only reality show about building an AI – and the stakes could not be higher. Four months. A new product. DevRev’s first-ever brand campaign. And one very big question: can we get everything ready in time? Meet Patrick Duffy, our Head of Creative, as he and the team try to figure out how to tell the world what DevRev is building. And why Computer, by DevRev, is different from every other AI out there. The problem: the product needs to be ready, too. Because there’s no point promising a whole new way of working if the product doesn’t deliver… a whole new way of working. Marketing, design, product, engineering, AI – everyone has to move as one. Season 2 takes you deep behind the scenes in the messy, ambitious, very human process of building Computer, and the brand around it. Watch Season 2 episode 1 now, and follow to get episode 2.

Everyone wants to build with AI. But when former @Verizon Business CEO Tami Erwin spoke with DevRev’s Head of GTM Development, Laura Fu, she offered a simpler place to start: buy the platform, then turn your attention to the work it should change. Pick one customer-facing workflow and redesign it for the way work could happen now. Know more: dvrv.ai/46WLKQJ

You shouldn’t need to know who to ask to get work done. 👉 Not when you’re trying to find the latest pricing guidance before a customer call. 👉 Not when a new hire needs the right policy, not five old versions of it. 👉 Not when Support is looking for the context behind a ticket that has already changed hands three times. The best help is the kind that gets your team unstuck before “quick question?” becomes a 14-message thread. Computer, by DevRev gives teams answers they can use – grounded in the knowledge your business already has. See what Computer can do: dvrv.ai/4hFAu11

The weekend was good. You logged off. You went outside. You completely forgot that Slack thread from 4:47 PM on Friday existed. Then Monday rolls around, and somehow you’re expected to remember: ➡️ What did we decide? ➡️ What needs to happen next? ➡️ Who owns what? ➡️ And where did that one important update disappear to? Cue the catch-up marathon 🏃‍♂️ Computer, by DevRev holds onto the context while you’re off being a person – so you can come back without spending half your Monday piecing Friday back together. So when Monday comes around, you’re not reconstructing Friday. You’re picking up where you left off. 🖲️ See how Computer keeps work moving → bit.ly/3TiHB69

Your AI strategy might be a very polished way to avoid making a decision. A new tool. A task force. A roadmap with lots of arrows. It can all look like progress. But at some point, someone has to ask the uncomfortable question: What are we actually trying to make better around here? That’s where Laura Fu, our Head of GTM Development, went with Tami Erwin – former CEO of Verizon Business and Fortune 100 board director. Tami’s take was refreshingly simple. We spend a lot of time talking about the technology, and not nearly enough time talking about what it’s meant to change. And until you do, “AI strategy” is mostly just a nice-looking slide. Watch the conversation 👇

The AI budget conversation usually starts with, “How much does it cost?” The more useful question is: What are we getting for it? An AI that requires more tokens to get to the right answer doesn’t just raise the bill. It makes every query, every workflow, and every new data source more expensive to scale. Computer by DevRev was built to be efficient where it matters – delivering accurate answers without treating your token budget like an all-you-can-eat buffet. Because the best AI investment should create room for more useful work, not more line items. See what Computer can do: dvrv.ai/4xR5mBj

A year ago, we launched Computer by DevRev. Today, it’s got a lot more context and zero crumbs in its keyboard :) Happy birthday, Computer 🎈

The worst kind of wrong answer is the one that sounds right. It slips into the deck, gets forwarded to the customer, and suddenly it is steering the conversation. And by the time someone spots it, the damage has already clocked in. Computer by DevRev is proven 48% more accurate than a leading frontier model – because enterprise AI should not need a babysitter. Think your AI can be trusted? Keep thinking or see the proof: dvrv.ai/4yvjKPT

Everyone’s talking about the next big model. Meanwhile, it’s Monday, and your team is trying to remember what was decided on Friday. ❗️Was the customer follow-up sent? ❗️Who owns the pitch deck? ❗️Where did those meeting notes go? And why does getting back up to speed feel like a job of its own? That’s the part Computer by DevRev helps with. It keeps the context, conversations, and work in one shared place, so the team can pick up where it left off – instead of spending Monday piecing everything back together. Because Mondays are hard enough. Your work shouldn’t have to start from scratch. See what Computer can take off your Monday plate: bit.ly/4yiSVxZ

We went to @scaler_official expecting to talk about AI, but we ended up talking about people. Turns out the two are hard to separate. Scaler spends its days teaching people to understand hard problems from the ground up. We spend ours building AI that helps teams get through those problems once they understand them. Different jobs, same mission. Somewhere in a room at Scaler, people are taking things apart just to see how they work. Companies spend fortunes trying to build that kind of curiosity into their culture. Scaler already has it, and it is the best possible foundation to build something with. Early days yet. We will let the work speak for itself. Thank you to everyone at Scaler for the warm welcome and the shared ambition!

At Global FinTech Fest in Mumbai, we’re bringing an AI with a better memory than the average group chat. In fintech, moving money can take seconds. Finding the answer when something goes wrong can still involve support, engineering, Slack, and a spreadsheet called 'FINAL_final_v7' Computer by DevRev keeps the thread – customer history, past investigations, product context, and what happened next – so the next person is not starting a fresh excavation every time a question lands. And it is the only AI proven 94.3% accurate – which makes GFF a pretty good place to put it on the spot. Bring us a real question at Booths J12–J15. We’ll be running live demos every 20 minutes, September 9–11. See you at GFF 👋

When someone who went from selling brick-sized cellphones to the C-suite at @Verizon tells you AI isn’t an IT project, you listen. Meet Tami Erwin – 35 years at Verizon, from rep to CEO; now a DevRev advisor and board member across four public companies. Her point is simple: leaders are getting too caught up in the technology. The real question is, what problem are you trying to solve? Because AI can change how work gets done. But it can’t replace judgment, trust, or the people who make a business work. Here's Tami’s take on leading through AI transformation: bit.ly/4ybRx0f

We are going to Gartner IT Symposium and Xpo on the Gold Coast this September, and we are bringing an awkward truth with us. Most enterprise AI has the memory of a goldfish. Actually, less 🥲 Sounds like a joke until you see the invoice. Ask it today, it nails it. Ask it tomorrow, blank stare. So it starts from scratch every time, slower and less accurate, and since you pay per go, that forgetting quietly lands on your bill. The fix is memory it can build on. Give it that and accuracy holds at 92 to 97 % even when your data grows 256 times bigger. One more thing worth knowing. Real AI does the work. It does not just log the ticket and hand the job back to a human. That is the whole show we are bringing to the Gold Coast from 14 to 16 September. Come find us at Booth 312. Bring your most cynical question. Those are the fun ones 🙌

Your customer asks one question. You open six tabs to answer it. The complaint's in Zendesk. The renewal they're sitting on is in HubSpot. The bug that started this is somewhere in Jira. And the thing you promised - "as discussed" - is in a Drive doc. Or Notion. You'll know when you find it. So you dig. And while you're reassembling this person out of six half-loaded tabs, they email again: "any update?" Computer by DevRev already read all of it – the ticket, the deal, the bug, the doc – so when you ask, the whole customer is right there. You answer them before they've finished refreshing their inbox. Everything connects. Computer makes sense of it. Know how: dvrv.ai/3SLvnml

We care a lot about what we build. We care just as much about who it's for. Last week, our team took that beyond the product – into a classroom and onto a start line. 🌱 First, Rise and Engineer 💻 – our CSR initiative with Empowerananya foundation and GITW in Bengaluru. We handed laptops to first-generation girl students and spent the afternoon talking careers, engineering, and backing yourself in rooms you weren't always invited into. Thanks to our speakers and panelists inspiring the next generation of women in tech. Then, Green BLR 2.0 🌳 – because impact isn't just what you build, it's what you leave behind. 69 DevRevers laced up at Cubbon Park to run for a greener Bengaluru, where every stride helped put a real sapling in real soil. Turns out our team ships trees as enthusiastically as it ships features. Empowering learners. Restoring canopies. Two very different finish lines, one belief: technology is at its best when it lifts the community around it. Here's to a week that starts with purpose. 💛

One wrong answer. Ten thousand customers. Ten thousand times wrong. That's how AI support fails. A calm, confident answer that happens to be wrong, handed to every customer who asks the same question. All day. All at once. And nobody sounds the alarm. A system that crashes gets fixed. A system that answers wrong with a straight face just keeps going. The damage never hits an error log. It hits customers who quietly stop believing you. The whole thing rides on what your agent actually knows. Feed it your real knowledge and all ten thousand answers hold. Leave gaps, and you'll spend the week apologizing for every one of them. Get the knowledge right, and being right at scale takes care of itself. But here's more on why AI gets confident answers wrong, and how to stop it: dvrv.ai/4hz4JHv

Every support team has lost an account that was frustrated for weeks before they walked. The discontent was sitting right there in their messages the whole time. Nobody clocked it in time. The first real signal? A renewal that quietly didn't close. CSAT surveys arrive weeks too late. A manager skimming a few tickets is reading a sample of the past. And a single "overall sentiment" score blurs every customer into one meaningless average. What you actually need: a live read on how every customer feels, right now, rolled up to the whole account, with enough explanation that you can act before the customer ever picks up the phone. Computer by DevRev scores every message as it lands – across chats and tickets – then rolls that signal up to the whole account. So when a customer moves from mildly annoyed to properly frustrated, you see it while there’s still time to act. And you see why – not just a score, but the issue behind it. That’s how the best saves happen: quietly, before a frustrated customer becomes a lost account. The account you almost lost? You’ll never even know you almost lost it. See how: dvrv.ai/4xXXJsL

The person shopping for a Tommy Hilfiger jacket isn't the person tracking a Flying Machine return. Treat them the same, and you lose them both. That's the reality Aviral Gupta, who leads Gen AI at Arvind Fashions Limited, set out to solve. With five iconic brands and shoppers across ticketing, WhatsApp, and the website, support had to get personal – and understand the context behind every conversation. So Arvind Fashions brought in Computer by DevRev. AI-native agents that talk to their data, pull it into one place, and show up the same way across every channel. The numbers backed it up: – Resolution times down 25 to 30% – First resolution times down – Around 25% of calls and emails deflected Aviral tells the whole story better than we can. Give it a watch 👇

It’s Monday morning. A customer writes in with a question your support team answered three times last week. The answer is there, but it's buried in a resolved ticket. It never made it to the knowledge base, so the AI agent can’t find it. Someone starts from scratch. Again. Someone already did the work. But because the answer is buried in a closed ticket, the next person can’t see it. However, Computer by DevRev, does. Running on Native Shared Memory, it spots the questions a person had to answer that the AI agent couldn’t. Then it writes the Q&A and puts it where the agent can use it next time. Next time the question comes in, the agent has the answer – and your team can stay focused on the work only they can do. The knowledge was always there. Now it stays. The best-kept knowledge base is the one nobody has to keep. See how it works: dvrv.ai/4z9hy1g

"We got a great rate on AI." Cool. It probably doesn't matter as much as you think. Two teams can pay the identical price per token and end up with completely different bills – because cost isn't set at the negotiating table, it's set in the architecture. And that's a product decision, not a procurement one. Demystifying Tokenomics is a straight-talking session on the real economics of AI: what drives the cost behind a token, how to tell a real advantage from a lock-in trap, and where the spend actually hides as you scale. Jeff Smith (DevRev) × Namrata Sarmah (Nam) (CPO Track®). Bring the skeptical questions. 📅 Monday, August 24 | 10:30–11:30 PM IST Register here: bit.ly/3TYdHnL

The Enterprise-Bench leaderboard has been live for a few weeks now, and the new standings tell a surprising story. Look at the top of the board: Computer takes the top three spots on three different models. The takeaway – it's the agent, not the model, that drives accuracy on real enterprise work. Enterprise-Bench is open and reproducible. Public tasks and dataset are on Harbor Hub, with the evaluation harness independently validated by Professor Alex Dimakis (University of California, Berkeley). New agents keep getting submitted, so the board keeps moving. But the question remains: who's brave enough to be benchmarked next? See the live leaderboard → dvrv.ai/4yXXE9L

"Seems like AI slop" is the internet's favorite reaction these days. So we made a version where Computer by DevRev earns the benefit of the doubt – it ships, it remembers, and it cites its sources. Meet the teammate who never forgets a thread: bit.ly/3TE7RrK

Enterprise-Bench shows the score. The trace shows the story. Tomorrow on DevRev Live, we're showing the Enterprise-Bench Community Leaderboard for the first time. We'll walk through the leaderboard together, dig into real traces from benchmark runs, and get into what we found testing open agent harnesses and open weight LLMs – the good, the surprising, and the stuff that made us go "huh." We've also got @smperng, CEO and Founder of JieGou.ai, joining us to share what the benchmark looks like from the community's side. Come see what the numbers say, and what the traces reveal. See you tomorrow.

Most enterprise AI looks incredible right up until the moment it has to work. That's not an accident. Demos are built to win. Clean data. A scripted task. "Success" defined before anyone hits go. Then it meets your real environment. Messy data. Shifting permissions. An answer that has to stay right as everything around it changes. That's where 95% of enterprise GenAI pilots fail. Yes, MIT counted. The answer is simple. Stop asking which AI is best. Ask if it's actually ready. Ready means precise, efficient, and safe when the data gets real. Everything else is theater. Here's the part that should worry every buyer. Right now, you can't tell the difference until you're already in production. Grateful to @TFinitive for running Jeff's piece. bit.ly/4yZXJcI

Somewhere in your org there's a list of things AI was supposed to handle by now. It's still a list. The models aren't the problem – they can handle it. What gets you is everything in between: testing that's more hope than process, zero visibility once it ships, and failures you only find out about secondhand. So the thing never makes it past "cool prototype," and the list doesn't move. On August 13, Shashank Sharan and Akhil Kintali are going inside Agent Studio to close that gap - live. We'll take an agent from a blank canvas to something running in production, and walk the full loop: build, test, deploy, observe. No code required to get started, and room to drop deeper when you want it. 30 minutes on the real thing. 10 for your hardest questions. Move something off that list. Save your spot: bit.ly/4gaAoxB

The best customer experiences never happen by accident. They're built on context, the kind that helps teams actually know their customers. That's exactly the conversation we're bringing to The Customer Show Exec Forum in Melbourne this week, alongside 150+ CX leaders who are rethinking what their teams can do with AI in the mix. So come find us at booth 112 and say hi. We'd love to show you how support teams are resolving faster, answering with more precision, and acting with real confidence – minus the busywork. See you at Flemington? Drop a comment or swing by – we'll bring the good conversation. 👋