.png)
One of the most important GTM moments in 2025 was when Docket closed a deal without anyone from our team on the first call.
A deal was sourced end-to-end by Alex, our AI agent on the website. It was equal parts exciting and slightly terrifying. Not because “AI booked a meeting” is a cool party trick, but because it validated a deeper shift we felt all year.
Buyers want the first real sales conversation before they ever meet a rep. And they want it on the website.
The Beginning: We Built an AI Sales Engineer
At the start of 2025, Docket was fundamentally an AI Sales Engineer.
Sellers depended on sales engineers to answer product questions—feature details, objection handling, competitive positioning, technical fit. We wanted to alleviate this dependency. So, we built an agent that could answer those questions with the same depth as your best sales engineer.
In just one year, 250,000+ questions were answered for our customers and thousands of salespeople relied on the AI agent for sales daily.
The feedback was specific, practical, and slightly annoyed they ever lived without it.
“Call prep used to take 20 minutes. Now the agent surfaces everything in 2 minutes.”
“We finally have uniform data in Salesforce. The agent fills fields humans always skipped.”
It worked. We have happy customers who continue to rely on this capability, and we're continuing to invest heavily in making the sales engineer agent even better. By no means are we taking our foot off the accelerator here.
But as 2025 progressed, we realized the opportunity was bigger. As AI capabilities evolved, so did our vision.
We asked ourselves: what if we didn't just help sellers answer product questions? What if we built agents that could execute the entire top-of-funnel motion autonomously?
The Expansion: Moving the First Sales Call From Zoom to Your Website
While seller behavior is still catching up to AI, buyer behavior has already shifted.
People now arrive at your website after doing their own research using tools like ChatGPT or Perplexity. They are habituated to instant answers. They do not want to wait three days for an inbound SDR to ask what their company does.
That’s when we made the sharper bet: the first sales conversation will happen with a digital salesperson, not a human rep.
We added the ability for agents to
- Replace inbound SDR queues with instant, product-accurate help.
- Qualify in real time and route to the right outcome.
- Book meetings directly, based on rules, not vibes.
- Do real discovery work—the kind that used to require a human.
In 2025, we proved it in production. First on ourselves (Alex sourced our own deals), then for customers. Conversion rates lifted, pipeline sped up, and meetings showed up qualified.
"Our agent-sourced leads convert at 4.7%, vs. 1.6% from traditional forms—almost 3× better."
"We're seeing 8–27 minute agent conversations—that's a full discovery call before we even talk to them."
"We set a rule: only push leads with >400 orders. The agent enforces it perfectly—our AEs trust the pipeline again."
The Architecture: What Makes a Digital Salesperson Different
Most traditional chatbots fail because they're built to answer questions, not orchestrate outcomes. But the buyer journey isn’t a conversation. It’s a sequence of decisions. And decisions require something more than pattern matching.
A digital salesperson is the combination of two things: Your best sales engineer's product knowledge + Your best seller's GTM process. And we built this on the Thinker-Responder pattern.
Thinker-responder = conversation + action
- Responder handles visitor questions (conversational)
- Thinker reasons and orchestrates tools: knowledge retrieval, slide/video delivery, qualification logic, CRM writes, calendar booking
Tool registry maps perfectly to buyer journey:
- Knowledge base → answers questions
- Verified knowledge → no hallucinations on pricing/security
- Slide retrieval → shows relevant deck pages
- Video → plays demo clips
- Calendaring → books meeting
- Human handoff → escalates when needed
We wrote a deep dive on this cognitive architecture here.
The Foundation: Why the Sales Knowledge Lake Made It Real
The foundation that makes all of this real is the Sales Knowledge Lake™.
It’s a centralized, governed repository that unifies product, pricing, security, slides, and call insights into one trusted source of truth. And it gives teams
- Faster answers: No scavenger hunts for the right slide.
- Auditability: Crucial for enterprise security reviews.
- Real-time trust: The website agent can handle deep product questions with grounded snippets, not hallucinations.
A digital salesperson without a Sales Knowledge Lake is just a confident hallucinator. With it? It becomes a reliable extension of your GTM motion.
The Evolution: From Single Agent to Multi-Agent Platform
The bigger shift happened when customers started asking questions we hadn’t anticipated:
- “Can we use this for partner enablement?”
- “Can we run different messaging for enterprise vs SMB?”
And here we had a moment of epiphany: we didn't need one super-agent. We needed a team of specialized agents, all sharing the same brain (the Sales Knowledge Lake).
We evolved into a multi-agent platform, allowing customers to deploy specialized agents across their GTM:
- Website concierge for inbound leads
- Sales agent for rep enablement
- Partner enablement specialists
- Enterprise vs. SMB specialists
- And several other use cases
All fed by the same Sales Knowledge Lake. All understanding your product and process the same way.
The fastest proof? Enterprise trust. By year end, we had 27 closed-won deals, with deployments ranging up to 1,800+ sellers at our largest customer.
The Breakthroughs: Features That Changed User Behavior
We shipped a lot this year, but the releases that mattered most were the ones that changed how teams work.
- Deep Thinking Mode (beta): Advanced reasoning for gap analysis and meeting prep. It felt like having an extra operator in the room.
- Real-time Voice Agents: Sub-2.5 second latency for live website conversations. Buyers could literally talk to your GTM motion.
- Live Assist: A sidebar that surfaces answers during the call.
- Time-to-Value: Customers went live fast. Some as early as 4 days. When you go from signature to production that quickly, implementation stops being the scary part of the deal.
The Validation: When the Market Starts Naming You
You know something is resonating when the market starts naming it for you. On calls, buyers described Docket as “ChatGPT for sales” or “Something that actually works.”
We didn't script that copy. It emerged.
External validation followed. Gartner named us a Cool Vendor in AI-Led Sales, and prospects referenced it directly. We held a 4.9 G2 rating all year.
At Gartner CSO, we generated 250+ leads in two days—proof that we weren’t just building another tool, but leveling up the company in real time.
The Hard Truths: Earning Trust Through Reliability
2025 wasn’t a straight line. One pilot walked away mid-engagement. That triggered an emergency debrief and hard questions about what “production-ready” actually means.
The answer wasn’t one feature. It was a thousand small commitments:
- Implementation in 2 weeks, not 2 months.
- Above 99% ingestion success.
- Agent-to-human handoff (because customers want smart triage, not just automation).
- SOC 2 controls ready before redlines.
The Engine: Designing a Team for Velocity
Speed is built into the system, not squeezed out of people.
We ended 2025 with a lean team across the US and Bangalore. The Bangalore office wasn’t a perk; it was a bet on high-bandwidth collaboration. Also yes, Bangalore traffic is now part of our go-to-market strategy.
“The ROI on this office is your happiness, the trust we build as a team, the unified understanding of our customers, and the speed of iteration we achieve.”
In December, we took the team to Wonderla for two days—not for trust falls, but to get past the polite version of the truth. That alignment is the real infrastructure under everything else.
The Future: Building a Platform of Digital Salespeople in 2026
2025 proved the thesis: digital salespeople can execute the first sales conversation better than humans could. In 2026, our goal is simple: become the essential platform for building and deploying digital salespeople.
Not just for website conversations. For every motion where a salesperson could be replaced by reasoning, autonomy, and speed.
The best closer this year wasn’t a person or a deck. It was the product doing real work before we joined the call. In 2026, we’re designing everything around that.