Technologies/Innovations Deployed

 2022: Employed by a supplier of pharmaceutical compliance services, we designed an AI supported process that reverse the traditional document processing/exception handling workflow, with advanced OCR capturing document images and AI recognizing the shape of each document prior to routing and batching.  This removed the need for a human “batcher” and several human intervention steps adjudicating unread documents. In the enhanced process, AI identified the document prior to the OCR read and resolving any handwriting, language, or position challenges, before routing the documents and output to the appropriate teams as CRM tasks. Ultimately, every inbound document (sample management, sample accountability, prescriber communication, etc.…) was sent to be “read” by the OCR system, interpreted with the support of trained algorithms, batched, and associated to a CRM task for an internal team to execute/acknowledge as needed, and then digitally filed for audit. The process enabled a company-wide dashboard of all customer-facing activity based on communication frequency, document type, and language "temperature".

Result: 82% reduction in documents re-handled/adjudicated.  Three Positions reassigned to value-added work.

2022: My team also designed an augmented reality/AI enabled process for sample locker inventories in regulated industries that reduced time spent at lockers, time spent training inventory specialists, and improved count accuracy while reducing the overall time spent reconciling. This was accomplished by allowing smart glasses to read the space, identify items, and prompt the user as necessary to correctly capture lots, SKUs, and quantities, reconciling the result in real time and directly prompting a recount, as necessary. If an issue arose that the user could not resolve with assistance from an intelligent agent, an in-house specialist joined the event remotely, in real time, and “saw” the same view as the inventory specialist. 

Anticipated Result: Significant reduction in time spent at lockers, greater time in field for pharmaceutical representatives, improvement in reconciliation counts, and >25% reduction in reconciliation follow up contacts transactions.

2021:  As a consultant for a distributor of medical literature, I directed the design and supported the implementation of a virtual pick wall for just-in-time matching of print-on-demand and variable print items to HCP samples and pre-kitted items selected by the warehouse. The system consisted of plain shelves and “smart” glasses supported by trained algorithms that learned the “shape” of printed items and used wearable technologies to project a virtual pick and put module, visually directing and confirming the placement of literature into bins which served as the second stage (another SKU) in the warehouse voice-directed pick. Integrating virtual put walls and two-stage picking had not been done by either system provider and required the internal development of significant “middleware” and validation system integration. 

Result: ~15% reduction in frieght and postage costs, co-mingling print on demand items with preprinted literature.  ~20% reduction in literature fulfillment labor due to elimination of batching, matching, and packing steps.  

2010: I recruited “fuzzy logic” experts from the telecommunication industry to improve identification and matching rates. This technology was ultimately replaced with “off the shelf” machine learning tools. The project was expanded to include the development of an easily configurable rules engine. At the time, the rules engine was built as the back end to a commercial ERP system, MS Dynamics AX.  The system persists to this day as the validation engine that allows the company to dominate it's space, selling a single view of customer activities that compliantly overrides selective rules or adjusts workflows for sample order validation and prescriber eligibility. Building that system just five years later would have required half the effort. 

Result: 18% improvement in matching rates, 1 day reduction in processing time for validation data identification and matching.

2024:  Although I have been involved in the development of many custom systems and interfaces, my bias is to leverage existing technologies wherever feasible.  To that end, I have sourced and implemented Manhattan Associates, Blue Yonder, and INFOR operations management systems, Oracle, Netsuite, and MS Dynamics ERP systems, Salesforce, Dendrite, Veeva, and Hubspot CRMs, Five9 Contact Center Management and Virtual Agents, and Pardot and Hubspot Marketing Automation Systems.

Results: Labor reductions in production of over 20%, Call Handling Time improvements of 18-25%, Shipment Cycle Time Reductions between 1-3 days.

2008: My team created a traditional data warehouse (before it was cool), utilizing SSRS and various presentation tools to move away from static reports. This data warehouse was an extension of our prescriber master database containing multiple identifiers and behavior based elements along with CRM transactions. At my next assignment, we re-created that resource on steriods, shifting from Cognos to Oracle and creating a dimensional “data mart” structure to support emerging (at the time) BI tools like Tableau and Domo. While supporting e-commerce operations, my team began adapting to the “data lake” concept, recognizing that our eCommerce ecosystem (supporting emerging brands on every type of system, including no system) was too chaotic for a traditional data mart approach. Today we are working with Snowflake to help us gain insights from our consumer and contractor relationships with (now) mundane AI assistance looking for insights in the shapes and patterns in the data.

 2005-2020: At the beginning of my career, we changed the way our space managed regulated distribution with RF directed picking and an in-house developed, paper driven, “put-to-label” system with sleds and “put tables” to compliantly break large 1-3 SKU batches of samples into discreet, auditable orders that were passed over a scale to confirm content. This concept is now mundane/outdated but it did not exist at the time we deployed it and had to be developed in-house by my team. Please note that at this time I was asked to assess the viability of robotic inserters at a competing business that was for sale and vetoed the acquisition due to the inflexibility of the technology. It’s not always about the coolest toys. 

At my next stop, we deployed RF-Picking to select batch orders which were loaded into a three-cart pick-to-light system consisting of shallow flow racks and pick lights which could be loaded by the warehouse and then positioned, and RF synced to the warehouse control system to direct the picker/packer to select the right order in batches for QC and labeling. This concept is now mundane/outdated but at the time we deployed it, the interfaces had to be developed by an offshore team.

At one of my most recent postings, we deployed voice-directed picking in a high-density sample pick module and a light directed put & pack wall to sort the batches into discreet orders to be labelled and shipped. The technology itself was mundane but the integration into legacy distribution software (AS400!) and CRM systems required a major development effort that delivered a level of structure, transparency and accountability that still does not exist anywhere else in that industry.  

Results:  Labor reductions up to 35%, reductions in order cycle times, increased order accuracy, improved employee retention.

 At every stop in my career, when presented with a similar problem to solve, my teams and I investigated emerging but stable technologies and applied them in ways that challenged their own developers to work within the constraints of regulated industries. I can provide similar experiences in call center technologies progressing from in-house developed “screen pops” to CRM integrations, to virtual assistants, and now in my current role, 100% AI based virtual agents.