What We’re Building

Problem statement (slide 1)

At Contour we’re building AI Deployment teams that replace professional services teams and deployment teams across B2b companies in midmarket and enterprise. To clarify, we sell to the VP of implementations and NOT customer success.

  • Nick Mehta pain - which can be extrapolated to other SAAS company
    • 0% ROI on them
    • 10M$ a year
    • Huge organizations that are oftentimes comparable to even the AE organizations
  • Samsara implementation team pain
    • Revenue on the table
    • Long tail of customers that don’t get serviced/trained properly

Vision (Slide 2)

  • Automated services segment completely, where professional services organizations and large implementation teams are completely agent powered
  • Full vision roadmap
    • Deployment intelligence layer for planning (Contour)
      • Includes creating knowledge base and dependency graphs for every ticket that needs to get done
    • PSA (we hook into this)
    • Agent operator (Contour)
      • Pulls tickets from PSA and executes them automating away percentages of work which is directly tied to manpower and revenue
    • Onboarding and walk through
      • agents walking through the environment they’ve setup and know every nuance of the custom implementation and why it’s important to the customer
      • We’re actually starting in this segment because the pain point of scaled segments allows us to penetrate implementation teams which is why our demo looks like it does

Show demo

Why Now (Slide 3)

  • Monitoring the space for the last few months and we believe computer use and long horizon agents are there
    • The team who’s going to do this is the most AI native team that can utilize the best in class models
  • Companies like Decagon/Sierra/Serval/Consol have models like this that are proven to be working
    • Implementation services are similar but in our eyes an even more bespoke problem which requires even better deterministic agent engineering which our team is capable of

Cut everything below should be questions only

Why are you starting on onboarding?

The product that we’ve built today, is the onboarding role of the implementation teams job. It’s a computer use agent that can operate on the site, speak to the user, and listen to the users input to guide them through the right workflows. The reason we’re starting here is because the wealth of user data that can be captured from walking users through each platform thousands of times. We’ll be able to better understand the specific bespoke usage patterns of each different type of user across every unique organization. This will help us build out the deterministic long horizon agents that are necessary for the AI implementation function that we’ve already started tackling with companies like Faire.

How are we thinking of building AI Implementation?

This is a huge problem space that’s extremely difficult. We’ve drawn inspiration from our friends at Decagon who have pioneered the idea of agent operating playbooks. Essentially we’ll own the implementation playbooks for every type of customer across every type of business.

The way we see it we’ll have

  1. A deep context graph of all the data sources that are relevant to implementation planning, this could include CRM integrations, data integrations, previous customer workflows, help center info and much more
    1. We’ll use this to build out dependency graphs where each node is essentially a task that a implementations engineer or someone will handle manually
    2. We’ll hook this into a PSA or task management software
  2. The implementation planning playbooks which instructs the agent on how to generate the dependency graphs from step 1 using the context graph. Each implementation plan will be layers of dependency graphs.
    1. The agent will work through different parts of the context graph and add tasks such as compliance access be required to run a data migration which is required to create a workflow that’s dependent on said data.
  3. The task building playbooks instruct the agent how to handle specific tasks in certain scenerios. The agent will work through different tasks that are pulled from the mentioned PSA or task management software, so there might be hundreds of these playbooks. These automate the human work which is directly correlated with cost for these professional services or implementation teams. The more of the operations playbooks we have the more work we can automate away
  4. The customer workflows playbook which instructs the agent on how to handle user interactions given the user input, user data and data of similar users on the platform. This will also query the context graph

We’ll also be providing tools that allow the user to sandbox these playbooks and understand exactly what action each agent is taking and why while also comparing it to previous human ran implementation workflows.

How are you different from Sable or WalkMe?

  • We’re replacing services (mic drop)
  • We’re going to penetrate the implementation teams specifically, we’re not going after customer success or DAPs
  • We will obviously win in the onboarding realm and create the best agent but it’s purely a penetration play to get into implementation teams where we’ll expand to building out the whole implementation services or professional service organization end to end flow.