Unlocking the Power of AI in RevOps
At Mops-Apalooza recently, I listened to Rey Ong, Vice President of Product at Openprise, talk about how to leverage AI in RevOps. Many people are excited about the promise of AI, as they have been for decades. Now that AI is increasingly available to at-the-keyboard staff, how should RevOps teams use it, and for what?
Rey first suggested having a framework for the evaluation of AI. There are three main types of AI: general AI LLMs, built-ins, and plug-ins.
- General AI tools such as HuggingFace’s custom LLMs, OpenAI’s GPT APIs, and Meta’s LLAMA are tools trained on generic data and can be further customized. B2B organizations are unlikely to have enough data to customize their own LLMs or will find it costly because the dataset needs to be tagged, which is labor-intensive.
- Built-in AI solutions such as OpenAI’s ChatGPT, Google’s Bard, and Salesforce’s Einstein and are trained on generic data with context to work only within a specific environment.
- Plug-ins or plug-and-play products with AI such as Gong, Marketo, and Zoom and are trained on more specific data and context with the aim to automate tasks, reduce manual work and increase efficiency.. These are the tools RevOps teams should explore first.
Is an AI-assisted workflow efficient?
Before using an AI assistant, ask if it would be efficient and truly help the business, your team, and your customers. Often in RevOps, each new tool or workflow needs an administrator or minder, which adds to the workload, creating less efficiency than imagined. Understanding your RevOps processes and its current inefficiencies will help you focus on where best to apply AI tools and whether it will increase overall efficiency.
Run through this checklist to evaluate overall efficiency:
- Is an administrator or workflow minder required? How much time do they need?
- Are salespeople doing more manual work (eg fact checking, editing, etc) to make this work effectively?
- Are content creators spending more time editing vs. shipping?
- Does AI-generated content from chatbots and support helpers create confusion among customers?
Security and risk with third-party AI platforms
How confident are you that the tool is secure and compliant for your environment and jurisdiction. It’s important to ask these hard questions because if you place proprietary data into the tools and they outsource their AI to a third party, how does your DPA cover this situation? Consider:
- You company’s compliance and risk procedures.what data you share with the AI tool.
- Whether the platform sends your query to another third-party platform to do the real processing. and whether your DPA covers this processing the potential risk to the organization and your customers if this data is leaked.
Operationalizing AI to Deliver ROI in RevOps
The next step is determining how to operationalize AI within your GTM operation and daily workflow. To me, this is always the most critical question for any new workflow or new tech. AI has to be applied to solve a real problem. Is there an inefficiency in the business that will be made more efficient with this AI tool? Would this new workflow cause other inefficiencies (and implications)?
For example, let’s look at possible efficiencies from a new workflow to pull Gong transcripts into Salesforce while generating a series of follow-up emails based on those transcripts to help sales follow up faster and more effectively. (thanks Michael Fan!).
- Sales reduces time to follow up with clear action items.
- Capture meeting transcripts and summaries to do more rapid sales manager feedback and give better context to other teams.
Potential friction and cost:
- Revenue Operations now needs a prompt engineer and/or workflow engineer to build and maintain this flow.
- Email copywriter needs to draft the frame or madlib format.
- Sales enablement required to train and evangelize this workflow. Does sales feel it helps them (what they feel may be a more powerful ROI than actual data)
- Human QA and sales editing needed.
Is the time savings greater than the resources to build and maintain this function? It very well may be!
According to Rey, the best area to apply AI is to automate rote activities using a plug and play products with AI in specific functions of your RevOps business process.
A great rote data example from Joao Antunes is for data normalization and enrichment from a human-entered field. Instead of picklists that may not be inclusive, he left the Job Title field in open text for any language. Then he can use an automated workflow to pass the field to an AI to categorize it to their preferred 20 picklist options. This is a great way to lower friction on Form Fill while letting automation handle data cleanup.
As we explore the tremendous potential for AI to be our assistants today and in the near term, it’s crucial to remember that AI relies on good, solid data as its foundation for making informed decisions and generating meaningful insights. Without high-quality data, AI algorithms can produce inaccurate results and unreliable predictions. How will you use it?