Last week marked 1 year since @joshdailey joined DocsBot as CGO. And man we've done a lot this year together! I started making a list for our annual review and it just kept going. With our small team, we've: - Shipped nearly 90 announced product updates - Sent 20+ email campaigns and published dozens of blog posts, product announcements, and guides - Published customer stories with StellarWP, Contractor Foreman, and GravityKit - Showed up at five conferences - Ran DocsBot Week with a week of launches, demos, videos, emails, and live builds - Rolled out a major pricing and packaging update with AI Credits, flexible add-ons, expanded BYOK options, and more flexibility for growing teams - Rebuilt onboarding with automatic website branding and source discovery, guided dashboard tours, and new free-trial signup paths - Launched our agency partner program with client handoff and ongoing management, and built out our affiliate resources - Revamped our SEO, rebuilt hundreds of existing pages, and added roughly 1,000 more pages across industry, integration, comparison, and educational content - Localized the marketing site into eight languages - Completed our SOC 2 Type II examination And then there's the product list 😅 - Deep Research: launched multi-step research across company knowledge, the web, and data analysis. Added question-history research to uncover bugs and documentation gaps, completion notifications, newer models, and usage-based access without a fixed job limit - Dashboard Operator: introduced an AI assistant that can configure bots, sources, and settings, with interactive questions to guide its work - MCP: launched our knowledge-search server, private-bot OAuth, question-history access, Admin MCP for managing DocsBot from other agents, and connectors for bringing external tools into DocsBot - Skills: introduced a plain-language Skill Builder for custom abilities and workflows, plus a library for reusing skills across bots - Agent API: upgraded reasoning, tool calls, multi-step workflows, and repeated searches so agents can keep working when the first search doesn't answer the question - AI Actions: added controlled web search, Stripe billing workflows, Calendly/Cal.com/TidyCal booking, and contextual action buttons that turn answers into useful next steps - Lead collection and CRM: added customizable lead capture, team notifications, and native delivery to HubSpot, Salesforce, Pipedrive, Zoho CRM, and HighLevel - Voice: launched website and phone Voice Agents in beta, added spoken input to chat, and introduced GPT Live conversations that can listen and speak at the same time - Messaging channels: introduced Facebook Messenger, WhatsApp, and Telegram integrations in beta - Slack: added multiple bots per workspace, channel defaults, @ mentions, streaming replies, and support for Actions, Skills, and MCP tools - Help Scout: improved AI replies with conversation logging, summaries, analytics, customer-property context, and controls for when the AI should stay quiet - Automation: expanded Zapier, added viaSocket, and introduced webhooks for leads, research, escalations, and ratings - Data Explorer: built visual exploration of training content and question history to spot topic clusters, gaps, overlapping content, and outliers - Search and infrastructure: migrated to a new vector database, added U.S./EU storage-region selection, semantic question-history search, and better ranked-chunk inspection - Document processing: added advanced parsing for complex PDFs and Office files, multi-file and ZIP uploads, individual file processing, and original-file downloads from citations - Knowledge organization: added Source Tags, filtering, bulk CSV management, custom source titles, and full-file previews - Knowledge sources: expanded Google Drive and Zendesk controls, and added Zoho Desk, Teamwork[.]com Desk, Salesforce cases, Box, Dropbox, GitHub, Bitbucket Cloud, and Scribe - Refresh efficiency: added scheduled YouTube playlist refreshes and cached unchanged OCR, image descriptions, and transcriptions to reduce repeat processing costs - Models: kept adding support from GPT-5.1 through GPT-6.1, expanded our model library and pricing calculator, and lowered AI Credit costs for several newer models - Dashboard and team management: redesigned navigation, added quick team switching, custom icons, per-bot permissions, metadata inspection, and printable reports - Chat experience: improved markdown, tables, formulas, diagrams, multilingual text, accessibility, and scrolling. Added dark mode, copy buttons, zoomable images, better branding controls, and live tool/reasoning activity - Internationalization: expanded the widget to 50 languages with automatic browser-language detection - Privacy and controls: added browser-side PII redaction and optional confirmation before opening external links - Developer flexibility: made it possible to update widget options without reloading or interrupting conversations - Billing convenience: added automatic paid-invoice emails and a separate accounting recipient Oh, and we built and open-sourced a physical DocsBot robot, gave it an iPhone and Apple Watch remote, and took it to a conference. Because of course the software wasn't enough 😂 Behind all of that were the customer calls, support conversations, demos, onboarding sessions, copy revisions, videos, launch plans, and follow-ups that turn something we built into something people actually use. I'm grateful to Josh for sharing that load and caring about where we're taking this thing. Having a partner to work through decisions with, challenge ideas, and help get projects across the finish line has made a big difference. Proud of what this little team shipped this year. Thanks for building it with me, Josh!

What if your LMS could have an AI tutor that understands your courses, videos, PDFs, website, and support content? 🤖 @ChrisBadgett talks with @uglyrobot & @joshdailey of @docsbotai about AI tutors, AI agents, voice assistants, automation, and LifterLMS.

Pushing GPT-LIve model support for our voice agents today

I shipped DocsBot Voice agents because some questions are faster to just say out loud. A customer stuck halfway through an install. A tech driving between jobs with grease on their hands. Someone who would rather talk than type out a support issue. They need an answer, quick So the bot you already trained gets two more front doors: voice via our widget, and a real phone line. It's the same agent you built. Same sources, same instructions, same tools. Voice presets for support, receptionist, and sales qualification. A full-power model when the call gets messy and needs cutting-edge reasoning. Mini when you want it faster and cheaper. Widget Voice and chat share one conversation, so a visitor can speak, then type, then speak again without starting over. And even in voice mode instant live transcription comes in with visual cards for all tools (billing, lead collection forms, and more). Phone Voice connects through a Twilio number or route via SIP. Leads get captured during the call. Transfers ring a live number during business hours, or an answering machine after, based on your rules. Wrote up how it works here: docsbot.ai/article/docsbot-v…

♥️ For my European friends and colleagues who haven't attended a PressConf (or formerly related Pressnomics) event, you will NOT want to miss this event. This event is where the "business of building the web" (with WordPress) sets the direction for the years that follow.