The inertia around leveraging large language models (LLMs) like ChatGPT is very exciting. The possibilities are astounding and uncomfortable all in the same conversation. Astounding because we have only just begun to imagine the possibilities; uncomfortable because we don’t yet understand what some of these possibilities mean to existing companies, operations, and teams.
As you begin to dip your toes into the world of artificial intelligence, you will make the most progress understanding what solutions make sense for you if you first know what types of business problems you may want to solve.
Consider the following scenarios in your company. Do these represent you?
We are a residential and commercial security company. We have a very large domestic and international customer base. We also have a very large product and services portfolio with multiple versions of devices, software, and service offerings, and a corporate intranet full of manuals, FAQs, processes, procedures, and industry referential materials. Hiring and training customer service representatives for this job is pretty tough due to the volume of information that must be learned in order to be timely and helpful to customers. And due to the speed and volume of ongoing change in our portfolios, it is a pretty crazy expectation for our customer service representatives to be fluent in everything, let alone helpful all of the time.
Is there a way for us to simplify what our customer service representatives must know, make information more easily available to them, and a way for us to guide them through difficult or complex support conversations with customers?
We are a very large property and casualty insurance company. Our underwriters are expected to attain and maintain one or more industry licenses and certifications to increase their expertise and enhance their credibility and reliability as professionals. In addition to ongoing industry training, we have personalized training programs unique to our company teaching them what we believe, what we pursue, how we work, and the business outcomes we desire for our own business solvency, industry regulation, and client delight. Our underwriters typically have multiple bookshelves full of three-ring binders containing underwriting guidelines, product specifications, coverages and limitations, processes, procedures, and algorithms to determine risk, pricing, and terms. And to top it all off, things change regularly.
Is there a way for us to simplify what an underwriter must know, what they must do, and how quickly they are able to come to a decision?
We are a business and technology consulting firm. All of our billable team members have utilization expectations, many of them have top line revenue and gross margin targets that must be met, and a few are additionally expected to hit EBITDA targets. Everyone is measured by our client net promoter score which is updated multiple times during an engagement and at the end. There are many moving pieces that must be taken into consideration for our employees to understand where they are in relation to what is expected of them, what it means in terms of remaining professional commitments for the fiscal year, and what it means to them personally in terms of professional development, promotion opportunities, and performance bonuses.
Is there a way for us to take all of these variables and create a reliable self-serve solution for our employees so that they can dynamically know their expectations for the year, what they have done to date, what remains to be done, and the associative performance, promotion, and bonus implications without always having to ask our People Operations team or Finance?
We are one of the largest residential and commercial manufacturing companies in the United States. Our client footprint is domestic and international. Our business growth is off-the-charts-amazing as a result of organic growth, but primarily due to our aggressive merger and acquisition team. We've purchased five companies in the last 36 months that make our product catalog unquestionably one of the most diverse portfolios available to commercial and residential builders in the industry.
The challenge we're having is that our product catalog is so large our sales distribution channels simply aren't keeping up with all of the options available during the sales process. We need something that will help us automatically understand the context of the build, the sale, and the builder and recommend the most probable, useful combinations of solutions to the sales distribution channels. Why? So that we minimize and eliminate missed sales opportunities due to lack of insight and/or oversights.
Our product development, legal, and marketing departments are having trouble looking for and realizing new opportunities that don't currently exist (opportunities in between existing products) due to the size of our growing product catalog. In other words, given we know what exists, we aren't doing a good job of identifying what doesn't exist, the patent opportunities, and the associative high-probability sales opportunities. Is there a way for us to prompt our folks on the highest probability, most useful sales while also helping our folks discover new products, services, and patent opportunities?
We are one of the primary companies responsible for providing hardware and software products and services for the federal government and military. In order to gain more business, we are required to fill out and submit over 500 RFPs per year and we anticipate that number to at least double in the next 18 to 24 months.
One of the problems we’d like to solve is to reduce how long it takes us to read, answer, and submit RFPs. Another problem we’d like to solve is that we don’t want to hire more people proportional to submitting more RFPs. We think it may make sense to create a situation whereby all of our RFPs are cataloged in such a way that we can check to see if we’ve already answered this question in a different RFP, if we’ve answered something like it, or if we’ve never answered it at all. This will give us the opportunity for reuse when possible, and build an ever-growing library of questions, project contexts, responses, and whether we won or lost those opportunities.
Is it possible for us to create a solution whereby whomever is filling out the current RFP can very simply query a system in plain English? For example, “Show me any questions that we have answered to date that are the same or similar to the following question from this new RFP: <insert question here>. Provide the date, RFP number, prospective client name, question number, question language, answer language, and the author. Tell me if we won or lost that RFP.
We are one of the few organizations hired as third-party auditors to verify that organizations are, or are not, in operational compliance with the agreed upon standard. We are hired by governing bodies and investors to provide insight pertinent to the purchase experience. Our auditors are held to a very high professional standard and are expected to follow our processes and procedures from beginning to end ensuring that the audit results represent reality and are not tainted or compromised by undisciplined or subjective behaviors of our auditors.
As such, we have a library of 1,600 process and procedure documents that guide what we audit, how we audit, how we verify observations, how we report observations, and what and how we recommend remediation thereafter. As you can imagine, our auditors have to know a very large body of knowledge and are expected to be precise.
Is there a way for us to make it easier for auditors to recognize obvious and potential deviations from expectation by looking up the pertinent process and procedure, comparing it to what they see, and getting recommendations on how to proceed? We’ve been documenting this stuff for many years and now have a very difficult time training people to competence. How can we increase the timeliness of our auditors, accuracy of their observations and recommendations, and ensure that we are building a body of knowledge versus merely relying on great people?
If you see yourself in any of these scenarios, or it reminds you of others not mentioned here, it may make sense for you to consider a large language model like ChatGPT, or whatever makes the most sense for your problem and solution set. Creating a situation whereby your folks can use regular English, ask anything, and get a structured response from all of your historical data reserves may be the next step for your organization.
Keep in mind as you journey on this path, you won’t have a problem finding a solution idea to your problem. More likely, you’ll need to work hard and decide what problem you really want to solve first.