FlowSmithy turns real work into trusted agent-callable workflows.

In 1968, Steve created an IBM 1401 simulator that allowed existing 1401 programs to run on IBM/360 hardware. The work helped preserve valuable business software during a platform transition, while taking advantage of newer system capability. The simulator was designed to fit naturally with the surrounding IBM/360 operating environment, including JCL-oriented production practices and demonstrated an early pattern in Steve's work: make legacy knowledge usable on newer platforms without forcing a disruptive rewrite.

It was marketed globally by Simpson-Sears, a major Canadian retail name and Steve's employer at the time.

At Canadian Tire Corporation around 1970, Steve built the CTC Source Control System to manage program source, copybooks, and JCL in a disciplined production environment. The system supported standards, controlled change, and automatic production of required back-level versions. It was used for years in a major corporate setting, solving the real operational problem of knowing exactly what source was current, what was historical and how to reproduce a prior production state.

In the Datacrown shared-processing environment, Steve developed an Interactive Optimizer to analyze and tune operational workload behavior. The tool focused on practical performance factors such as JCL structure, file blocking, indexing and processing patterns. Its purpose was not just to make one job faster, but to turn expert performance tuning into a repeatable process that could be inspected, adjusted and applied systematically.

Its use gave Datacrown a distinct competitive advantage in securing more shared processing workloads.

Steve designed many-to-many word-processing conversion technology for systems such as AES, MICOM, DEC, Vydec, IBM DisplayWriter, Xerox, Wang and WordPerfect. Rather than writing a separate converter for every possible pair, the approach translated documents through a normalized internal token/data stream. This made the conversion system extensible, practical and economical, and it ran on modest S-100 CP/M systems using a PL/I compiler and 8086 assembly support. The implementation included a log that identified points in the document that needed manual inspection and/or correction. That was a unique feature among the conversion tools available at the time. The design anticipated ideas later seen in common interchange formats: separate the meaning of a document from the quirks of each vendor's file format.

Solutions Canada Inc. was created in 1983 as a result of this capability. It gave word-processing vendors such as IBM, Xerox and Wang a company to point to for solving the problem of migrating old-format documents to new vendor formats. Many of their clients, including McDonald's, TTC and CIBC, became Solutions Canada clients.

In 1988 Steve created a Remote Operator approach for IBM 4680 point-of-sale systems that enabled home-office employees to connect to a store POS system as if they were there. To achieve this goal, Steve had to understand and interface with a highly protected operating system not designed for conventional remote control. By analyzing how the system handled operator interaction, he found a practical path to intercept and replace command behavior in a way that enabled remote operational support. The work reflected a recurring strength: understanding a constrained technical environment deeply enough to create a usable control layer above it.

It was marketed by IBM to its global retail-market clients, including Walmart, CVS, The Gap and Karstadt. IBM also produced a Redbook that described its use and implementation.

In the late 1990s, Steve built a Y2K inference tool that identified date-related fields by examining how data was used, moved, compared, written and read across programs and files. The tool did not rely only on field names, because important date semantics are often hidden behind local naming conventions or inherited data structures. By propagating date meaning through the more reliable evidence of data flow and usage, it helped large organizations find Year 2000 risk more intelligently than simple text scans could.

It was used in PricewaterhouseCoopers client work to help with Y2K migrations.

FlowSmithy brings Steve's long-running pattern forward into the AI-agent era. It is a workbench and execution framework for turning operational know-how into structured workflows with declared inputs, controlled execution, status, logs and concise results. The current system includes a configurable runner, workflow configuration tooling, and multiple execution paths including Tcl, Java and PowerShell. The next natural layer is an MCP-compatible headless interface that allows AI automation agents to discover approved workflows, supply validated inputs, run them and receive structured results while humans retain visibility and control.