After more than a year of development, Sevanta is proud to announce Raven, a powerful AI assistant that is fully integrated into the Sevanta experience. Sevanta Dealflow can continue to work as it always has, but clients who choose to enable Raven get access to some astounding functionality at no additional cost.*
Raven starts as an unobtrusive chat input box in the lower right corner, which can expand into an AI chat sidebar. Out of the box, Raven has access to our documentation library, so it can answer questions you have about Sevanta, but more importantly, it has access to all the data in your dealflow system, including files and emails. So Raven can answer questions, compare deals, summarize histories, and more, but you can also ask it to make changes to the data in your system, including creating new deals, adding comments, setting reminders, or making a deal inactive.
Raven can also search the web if it needs outside information to answer your questions, can access specific URLs you give it (like to fetch information from a company's website to help fill out a deal record), and can send emails to you.
And that's just out of the box.
You can optionally connect Raven to other systems, including your email and calendar, so when you ask it a question, it can selectively retrieve relevant emails from your Gmail or Outlook that might not exist inside Sevanta.
Connecting your email dramatically increases the power of Raven; for example, you could make a daily automation that looks to see if there are new deals you forgot to log, and creates them in Sevanta automatically for you.
You can also ask it to draft emails for you, but we have Raven on a short leash: it's not allowed to send emails without you in the loop. You get to review and edit the emails in your drafts folder first, and then can choose if and when to hit send.
While Raven has built-in support for a lot of common platforms, and we'll continue to add more, you can also connect any MCP server to it, and then run automations that span all the tools you give it access to.
The idea is that Raven provides the scaffolding that can connect everything together, so that you can just focus on the business problems you want to solve, and you just have to describe what you want in natural language. Then automations can run on whatever schedule you want, also described in natural language.
But wait, there's more!
If you connect your calendar, you can enable the Notetaker to join your meetings. It's a bot that can join any major platform, including Zoom, Microsoft, Google, and Slack, takes a transcript, and then adds summaries to the corresponding deal in Sevanta automatically. If you're used to some other notetaker app, of course you can keep using that, but the automatic integration with Sevanta with a single click of a toggle is pretty amazing.
Also, while you can interact with Raven in the sidebar, or in a full tab, you can also message it via lots of major platforms, including Whatsapp, Telegram, Slack, Signal, and even basic SMS. When you talk to Raven using the same app you use to talk with your colleagues -- and you even ask a lot of the same types of questions and ask it to make changes or draft emails the same way you might ask a personal assistant -- you start to realize that Raven really can become a virtual member of your team.
FAQs
How much does Raven cost?
Every user gets $10/month in token credits included with their Sevanta subscription; given that a typical simple question costs 5-8 cents, you can do quite a bit of testing without exceeding that monthly credit. However, if you use complex prompts that pull a lot of data, a single run could cost $3-4; and if you're running it daily as an automation, you could exceed the built-in credits quickly. If you drop below a $5 balance, Raven will start warning you that you're low, and will ask you to enroll a credit card to cover "extra usage". If you do, then it will charge you $15 every time you drop below $5 (those trigger points are configurable by you, but are the current defaults and minimums). This approach thus does not commit you to any particular spend; you'll just be charged based on how heavily you use Raven, and for light usage you won't need to enroll a credit card at all.
What LLM does Raven use under the hood?
Currently Raven uses Anthropic's Claude via Amazon's Bedrock service. We're agnostic about that, and if in the future other providers offer better quality, speed, or value, then we may change that, but for now we chose Claude for its quality, and we chose Bedrock to avoid using an API directly to a foundation model developer. Bedrock is an "ephemeral" service that doesn't store or log inputs and thus can't use inputs to train models or share with 3rd parties. When you see your chat history in Raven, that is stored only inside Raven -- nowhere else -- and if you delete a chat in Raven, that properly deletes all data associated with it.
What about privacy?
As mentioned above, we chose Bedrock as a mechanism to minimize 3rd-party risks and to ensure your data can't be used for training or other purposes by an LLM developer (and Sevanta does not perform our own training or fine-tuning). As for Sevanta, your data in Raven is subject to the same confidentiality provisions we already have in place with you per the subscription agreement. We have access to chat history and other internal data for the purposes of providing support and fixing issues, but your data will be used for no other purposes whatsoever.
Also, keep in mind one key difference between Raven and Sevanta Dealflow: your dealflow system is a shared workspace for your team, so data there can be seen by your colleagues, but your relationship with Raven is just with you; e.g. if you connect your email to Raven, then Raven can only use that connection when interacting with you. And your chat history is not visible to any of your colleagues, but if you ask Raven to add information to Sevanta Dealflow, then that information will be visible by your team.
Building Raven seems like an enormous undertaking; why did Sevanta choose this route?
Indeed, it took over a year to get Raven to the place it's at today! But the alternative, which a lot of other CRM and SaaS providers are taking, is to simply open an MCP interface to their platforms. The problem we have with that is that we'd have no ability to provide any sort of support: if your AI gives you bad output, we'd have no ablity to diagnose what went wrong and no ability to fix it. With Raven, we control the full stack: the user interface, the "harness" (logic loop, orchestration, etc.), the API connections to various services, and everything in between; so if you click the "thumbs-down" button in a chat, we can reach out to you to help, whether it's just a matter of suggesting tweaks to your prompt, changing the API endpoints or how those are presented to the LLM, or various aspects of the planning and logic process in Raven. We're committed to supporting you with Raven to the same high level of excellence that we support the rest of the Sevanta offering, and that's something we see nobody else in AI doing today.
