Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

The future of B2B buying is personalized: Using AI to halt friction

When built on a curated library of data and constructed by a forward-thinking engineering team, AI models can improve customer and provider outcomes and relationships.

[Images: Blue Planet Studio/Adobe Stock]

Profile photo of Sergei Dubograev

BY Sergei Dubograev

AI is like a movie star—it has a major role to play but won’t be in every scene. We are on the cusp of the biggest cultural technology shift since the introduction of the internet and mobile phones, so it may be tempting for businesses to think of AI in terms of widespread deployment rather than more targeted approaches. But when it comes to improving the business-to-business pipeline and processes, we have found that leveraging AI is about quality, not quantity. 

The focus should be on identifying the use cases for AI that have the biggest impact. It is a lesson for everyone as businesses come to grips with the potential of generative AI to create a more personalized product with its contextual learning capacity.

But if we want a more customized output, customization of AI itself is needed. For companies in the B2B space, that requires building their own models trained on both open-source and private data. This will allow them to analyze buyer personas and even anticipate customer needs before connecting them with the resources and services that solve their problems. 

We call this the next level of helpfulness, and it can naturally evolve into loyal customer relationships.

BUILD IT AND THEY WILL COME

As more big organizations make language models open source, the availability of vast data sets is going to change the AI industry. It’s a village approach to mutual growth as smaller businesses access advanced language capabilities without major investments. 

Training AI on public and private data is the best of both worlds because it allows clean proprietary data to be leveraged for an unprecedented level of customer targeting. So my advice is to conduct market research into customer pain points as you identify your biggest opportunities. Once you know where the friction lies, consider the best use cases—not the most use cases—to decide what features the model should include. 

But fair warning: Developers operating in a vacuum often come up with solutions that do not match business needs. Take a cross-functional approach to putting the right team and the right tech in place to get the building process underway. Involve your marketing teams, sales personnel, and other departments on the front end early in the process so the data scientists will operate within the guardrails and toward the business objectives you have already established. 

DATA IS YOUR FRIEND—TREAT IT WELL

Any AI model is only as good as the data it was trained on. There are two ways to collect quality data before building a language model so that it can be accurately trained. 

One is to track user behavior and create personas to evolve user profiles and gain insight into the target audience. Businesses seeking assistance with SEO, for instance, are likely to have similar needs based on tracking their search queries and web behavior. The second is to train the model based on the organic responses of buyers in surveys, reviews, and chatbot interactions. 

In order to catalog all of these connections and determine what will be included in the training set, analyze and correct the structure of the data by:

  • Spending a lot of time crafting and refining with prompts
  • Extracting only relevant data and disregarding the rest
  • Validating contextual-search learning data points
  • Testing the effectiveness of the algorithms
  • Continuously refining the model with weekly and sometimes daily updates

Also, pay attention to buyer feedback to see if they received what they needed from the interaction. On the supply side, communicate with vendors. Was the lead successful, and if not, why not? Ensuring the data is categorized correctly and meets the goals of what you want AI to achieve is critical to building a strong language model. 

PUTTING THE MODEL INTO ACTION

Now that the AI model has been optimized for use, it is able to remove friction from customer interactions. Consider the case study of a CEO who needs development services but doesn’t know any programming languages or even the most appropriate service. We can train our AI to connect businesses directly with providers, so it will focus on the CEO’s business challenges rather than requiring the exec to ask technical questions.

In the past, AI was limited by its reliance on rule-based systems and shallow learning algorithms, which constrained its ability to effectively process and utilize vast amounts of data. However, with advancements in deep learning and embedding techniques, AI has undergone a remarkable transformation. It has become significantly more powerful and efficient in leveraging data, enabling it to uncover intricate patterns, extract meaningful insights, and make more accurate predictions than ever before. But this new type of AI language model will search reviews and find successful projects so the CEO will know how the final project would end up looking with the right provider. 

It effectively bridges the knowledge gap through its capacity for contextual learning and not only figures out the questions the CEO should have asked, but answers them, too. Using an AI model like this allows leaders to manage expectations and be one step ahead of the needs of users, even when they cannot properly articulate those needs themselves.

BE WHERE THE CUSTOMER IS GOING TO BE

Generative AI has a place in all of the potential additional features of a language model, from research to choosing a vendor to completing a project to planning the next one. It can also personalize the options of project management, contract, and communications tools for a total solution.

When built on a curated library of data and constructed by a forward-thinking engineering team, AI models can improve customer and provider outcomes and relationships. The capacity of language models to learn patterns and context can help B2B businesses anticipate where the customer will be, even before they show up. In other words, the more we train the model, the smarter it becomes. And the scariest thing about AI is its limitlessness. 


Sergei Dubograev is Vice President of Development at Clutch.