As a sales rep working with large enterprise accounts, one of the things I struggled with was understanding the organization structure of some of these large complex companies. I remember having a folder structure in Outlook/Google Drive that had the account, the division within the account, the department or team under the division. Then each time I met a new person, I'd try and map them to one of the teams or if it was a new team, I'd create a new folder.
This team level understanding of the account was extremely useful for me as I could connect the dots with use cases relevant to a department, people I knew within a department etc. and thus show up with credibility.
As part of my research into account planning tools and approaches with GenAI, I wanted to see if there was another way to do this. The ideal way is for the customer to share their org structure but that is not always possible.
So I embarked on a journey to do this using job information. Surprisingly, job postings tend to have a lot of detail around the division, groups, their priorities and what they are expecting this role to do.
Here is an example for a job at Morgan Stanley -
Legal & Compliance Division (LCD) comprises of Legal, Compliance, Global Financial Crimes and Regulatory Relations.
The Legal Department provides guidance, requirements, and procedures for understanding and complying with the laws, regulations and Firm policies that apply to our businesses.
The Global Compliance Department identifies applicable Compliance Obligations and maintains a Firmwide Compliance Risk management program, including Compliance Risks that transcend business lines, legal entities and jurisdictions of operation.............
It also talks about what each department / team does, what projects / priorities they have and sometimes what success for each role looks like and even what technology is used.
Obviously, if the company is not hiring then this is not very useful. That being said, often if a company is not hiring, then they might not be buying technologies either.
The process I'd follow before GenAI is something like what I have listed out below.
I'd first try and find a list of jobs for a particular Division using LinkedIn or Google. From this I'd have a list of jobs. Then I would open each job 1 by 1 and read it and extract all the information - Division, Department, Team, what each department and team does, what their strategic goals/projects are, what challenges they have, what key success metrics they are looking at, what technologies they need. Then I would repeat this for the next job.
Here is some simple math - 60 jobs x 30 mins per job = 30 hours.
Needless to say, I would not do this and I would lose out on some very useful account information.
So I set out to do this with GenAI. What I quickly realized is that there is no one technology that can do this. Secondly, I could not completely eliminate manual effort, but even with a combination of GenAI + manual effort I was able to save a ton of time.
The process I followed was to use Clay to generate the list of jobs. This is super super easy and took about 5 mins to get a list of jobs. I looked for Jobs at Morgan Stanley in Legal and Compliance. It came back with 60 jobs.
Now Clay has the ability to scrape websites (and 3 different ways to do this), so my first attempt was to use Clay to go to the job listing URL and scrape the job description. What I wanted to do was to then collect all the job descriptions and put it into one document and use GenAI against it.
While Clay can scrape websites it tends to store the output in a cell in a table and the job description has too much text.
I could use Claygent their GenAI plugin to read the page and summarize it, but then I was losing out a lot of the details I needed to answer all the questions.
In the end, I ended up copying and pasting the job descriptions into one big PDF file. This took some time but overall was about 15 minutes.
https://docs.google.com/document/d/1ivs1-G2LhLcQeSs5oQDxxBWLqqpgVrXaVb5N_RHgSl4/edit?usp=sharing
Now that I had a PDF with 60 job listings, I put this into ChatGPT and asked it my series of questions via a series of prompts. This took about 5 minutes and I had all my answers.
With GenAI + Manual Effort = 30 mins.
So what would have taken 30 hours ended up taking 30 mins. The new process is below.
I had posted earlier about my journey with GenAI and how there were different GenAI solutions that you need to pick from depending on what you are trying to do.
While this need falls under Account Research, I realized that the POINT SOLUTIONS out there did not have this out of the box, so that was not an option. FEATURES IN EXISTING PRODUCTS like LinkedIn / Zoominfo were not that helpful either. So I ended up on the 'Custom build' side of the equation and ended up with a combination of LLMs (ChatGPT) and Agent Builders (Clay).
What I am realizing is that to solve problems, you might need a combination of solutions and you have to be enterprising. While the end solution is quick, it does take time to figure these things out. I would say that I spent over 10 hours researching options, playing with different tools and slowly formalizing an approach to tackle this.
The output (see sampling of Projects, Challenges, Success Metrics) are way better than anything I would have created as part of my account research. So it is definitely worth it, once you get to the end.
