When a problem presents no obvious path forward, the typical corporate response involves either throwing money at it or waiting for a sudden stroke of genius. Finith Jernigan, Ph.D., prefers a much more systematic routine. He treats an impossible challenge as a basic puzzle that simply needs dismantling. By visualizing the final objective first, leaders can work backward to uncover tiny, manageable steps. After mapping out those distinct pieces, modern software tools can search completely unrelated industries to see how parallel issues were handled previously. The approach takes human intuition out of isolated silos and relies on pattern recognition across massive amounts of data.
Working Backward to Find Answers
Staring down a blank whiteboard paralyzes even the most experienced teams. Jernigan knows this feeling well and points out that getting stuck is a very literal part of the process. “Seemingly impossible to me means you’re starting with a problem you have no idea how to solve and no idea even what the first step is,” he explains. Rather than attempting a wild guess at the onset, his method involves shifting all focus to the very end of the project. A team can plot a course starting from a complete solution, tracing the logic back to 80 percent done, and continuing down until they reach their current position. Defining the actual steps needed to travel between those points is often the most critical part of the job. “If you can break this large goal that seems impossible down into five of those types of steps, you can really link them together in manageable bites,” he notes.
Finding solutions to those isolated micro-steps rarely requires inventing something brand new. Somewhere in the world, another field has likely tackled an identical hurdle under a totally different name. Jernigan relies heavily on searching beyond his immediate sector to track down odd or creative fixes that might translate well. “That’s where AI really helps me, seeing examples you can find through searching for other people who have solved this and how they did it,” he points out. Reviewing those outside strategies resembles a traditional pitch meeting where multiple angles get tossed onto the table. Jernigan specifically compares the dynamic to watching a television show like Mad Men, where an executive listens to 10 different marketing ideas before selecting the perfect one to execute. The only actual difference today is that digital software agents are now delivering the pitches instead of human marketing teams.
Rethinking Traditional Research Habits
This perspective draws heavily on an early desire to shake up established lab routines. Basic scientific research generally operates in a straight line, demanding researchers pick an interesting target protein and brainstorm which patient groups might experience relief from it. Jernigan decided to flip the map entirely. He found greater value in reading off-label clinical results to find the anomalies first. Seeking out positive patient outcomes opens up fresh ways to look at basic biological mechanics. “What patients have responded to what particular types of therapy, and is there some insight you can get that leads you all the way back to basic research?” he asks. Doing this manually used to dictate endless trips to a medical library. Tracking down a handful of similar private practice observations meant digging through thousands of disconnected files. Now, smart applications comb through medical archives and sort similar incidents in a fraction of the time. “If you can find multiple physicians with the same result, that’s where the AI tools really come in, because then you can scan everything, ask very specific questions, and find the reports that are very similar to each other,” he explains.
Delivering Results Through Autonomy
Software capabilities are currently scaling past basic internet queries into autonomous action. Agentic AI completely alters how teams explore possible project roads because humans no longer have to direct every individual task manually. Setting up a string of connected digital helpers allows complex workflows to happen essentially overnight. A user can tell one bot to locate 10 interesting solutions to a problem, while instructing a second bot to generate a detailed report on each concept. Once the data gathers into one place, a third agent distills everything down into a tight final document. “Suddenly this agent comes back to you and says, ‘Well, I’ve completed all 10. What else do you want me to do?’ That’s like having a whole research team of 100 people 20 years ago working for you,” Jernigan observes. The net result essentially eliminates the danger of overlooking a workable strategy just because the human staff ran out of clock hours.
While digital logic feels highly refined at the moment, interacting with the real world is the obvious next frontier. Bringing software logic into a physical room will change how businesses handle everything from manufacturing to chemistry. Scientific testing regularly suffers from consistency bottlenecks because some lab results are notoriously difficult to replicate without intense niche expertise. Finding ways to transfer specialized knowledge out of human hands and into automated machinery could level the playing field entirely. “If you start getting into robotics at the highest level, you can train the robot to do those very difficult experiments. Suddenly, anybody can do them with the right recipe,” Jernigan says. The moment machines can physically understand and manipulate their surroundings based on data queries, tackling an impossible problem will look exactly like running a piece of everyday software.
Follow Finith Jernigan, Ph.D. on LinkedIn for more insights on systematic problem-solving, agentic AI, and leveraging technology to solve complex scientific challenges.
