Context creates better AI outcomes
Far out, there’s so much bullshit around AI right now. Marketing hype, vaporware, and named “agents” you can “hire” to “replace your team.” We’re told that everyone will soon be using Claude as their primary interface to do everything. I’m not so sure.
We work with a lot of large, global companies. Most of them don’t give us publicity rights, but you can see some of them on our customers page. Spending time with these teams has convinced me the popular story about AI is missing something important.
Generic tools, generic results
As people have experimented with general-purpose AI tools and agent platforms , something I’ve heard again and again is that the results they’re getting are generic and underwhelming. That makes sense. Base AI models are trained on publicly available data and they need a lot of context to do a good job at a specific task.
Take just one example: ‘synthetic users’, or ‘synthetic personas’. I know many teams are trying to create these with generic tools and they’re underwhelmed by how shallow the output is. One of our customers called their synthetic customer persona a “caricature.”
If you think about it for five seconds, it makes sense. Consider hiring a new employee. It takes time to onboard someone because they need to learn about your business, market, customers, users, and how the business communicates before they can do great work.
The same applies to ‘synthetic users’. If you create these with generic AI models, they’ll produce generic AI responses. I believe that as long as businesses try to use generic AI platforms to accomplish specific tasks, they’ll guarantee generic outcomes.
Aaron Levie, the CEO of Box, put it well:
We are entering the era of context. The teams and companies that can accumulate and best utilize context will drive the greatest productivity and highest output.
— Aaron Levie
To increase the quality of the work, we need to provide context.
Dovetail helps teams provide context to AI
Today in our Sun’s Out launch, we launched a host of new features in Dovetail all designed and built around this one idea: how can we provide AI capabilities that do a good job at specific tasks because the model has the right context in the form of private business data, customer interaction data, and insights?
Let’s go back to synthetic users. One of our new features is ‘digital twins’ in Dovetail, where you can create an agent that mimics a specific customer (e.g. Volvo), customer segment (e.g. Free Users, Enterprise), persona, role, or archetype.
Because these digital twins are powered by Dovetail, they can draw on a rich repository of real interviews, sales calls, reviews, CSAT responses, and support tickets—all indexed, cached, enriched with metadata, and presented in a format the model can process quickly and reliably. The data resides in the same place as the AI.
Context is the difference between using a generic AI and hoping for specific outcomes, versus a purpose-built intelligence platform with data, apps, and context baked-in.
What we’ve built
I’m super proud of the team for all of their hard work over the past six months bringing digital twins in Dovetail to life along with several other awesome new capabilities that help teams get better results from AI when working with customer insights.
Agents, now generally available. Autonomous teammates built on a simple trigger, skill, and tool model. Start one on a schedule, on demand, from a digital twin, via an event like a new call in Dovetail, or an external event like a new opportunity in Salesforce. Agents in Dovetail sharpen their own skills across runs, trace every output back to its source, ask for approval before any write action, and never act beyond your own permissions.
Channels 2.0. Automatic feedback classification that turns raw, qualitative customer touch points into revenue-weighted ideas. It ingests from 30 or more sources including Intercom, Salesforce, App Store, Google Play Store, and Gong; proactively surfaces new ideas and shows you who’s affected and how much ARR is at stake; then allows you to move that context and evidence to Linear, Jira, Claude Code, or Cursor in one click.
Learn more about Channels 2.0 →
Global and local context. You can now add your strategy, brand, tone of voice, and compliance guidelines once at the workspace level, or to a single project or channel, so that every AI feature has the context it needs to improve accuracy and reliability.
More ways in and out. 10 new data sources, including Qualtrics, Snowflake, HubSpot, Pendo, and ServiceNow, plus first-party MCP connectors that push Dovetail’s intelligence into the tools teams already live in like Slack, Linear, Notion, Figma, and Salesforce.
External MCP in chat and agents. Connect external MCPs from Hex, Canva, Notion, Salesforce, Linear, and more to Dovetail’s chat and agents features. Pull in context from those tools (e.g. Hex analytics) and push outputs back to them (e.g. Canva deck).
Governance and compliance. We’re now ISO 42001 certified for responsible AI and have launched AI redaction, custom data retention, and full version history for docs, ensuring Dovetail provides the controls over AI that enterprise teams need to sign off.
Learn more about everything we’ve announced here: