01
Remain human
Judgment, accountability, relationships, ethics, and high-consequence decisions.
AI transformation advisory
From AI possibility to successful adoption.
Dividing Line Group is a firm of Senior Operators helping organizations determine where AI creates real value, architect the environment required to capture it safely, select the right technologies, and move them into production.
The problem behind the technology
Most organizations do not have an AI-access problem. They have an AI-transformation problem.
Pilots proliferate. Vendors promise transformation. Internal data is fragmented. Sensitive IP creates legitimate constraints. Teams disagree on platform versus point solutions. Governance arrives late. And promising tools often remain outside the workflows where value is actually created.
The dividing line
Decide what AI should change—and what it should not.
Every workflow has a dividing line between work that should remain human, work that should be AI-augmented, and work that can be automated. We help clients draw that line deliberately, then design the technology and operating environment around it.
02
AI-augmented
Research, analysis, design support, drafting, knowledge access, and decision support.
03
Automated
Repetitive work, routine handoffs, standard synthesis, and low-value administration.
How we work
Five connected stages from opportunity to adoption.
The framework is deliberately continuous. Architecture informs vendor selection. Selection anticipates implementation. Governance is designed in rather than bolted on later.
01 · Assess
Start with the work—not the tool.
Map workflows, pain points, existing pilots, data dependencies, and where human judgment remains essential.
Client output
- Prioritized AI opportunity map
- Readiness and risk assessment
- Value measures and success criteria
02 · Architect
Design the environment before selecting vendors.
Define AI experiences, enterprise knowledge, models, orchestration, APIs, cloud/on-prem choices, identity, and the boundary around sensitive data and IP.
Client output
- Target AI architecture
- Data/IP and security boundary
- Integration and human-control requirements
03 · Evaluate & Select
Choose objectively.
Build requirements and evaluate enterprise platforms, specialist tools, models, and build-versus-buy options without vendor allegiance.
Client output
- Evaluation framework / RFP
- Technical, economic, and risk comparison
- Recommendation with explicit tradeoffs
04 · Implement
A pilot is not success. Production adoption is.
Program-manage the transformation around the technology: parallel operation, acceptance criteria, integration, training, metrics, and executive decision gates.
Client output
- Implementation roadmap
- Acceptance and sign-off criteria
- Adoption metrics and operating cadence
05 · Govern
Build controls into the operating environment.
Establish practical security, privacy, legal, compliance, model-risk, and performance controls that enable controlled speed.
Client output
- Governance framework
- Roles, controls, and escalation paths
- Ongoing monitoring and review model
What this looks like in practice
Specific AI-transformation problems—not generic business consulting.
DLG starts with the client's AI decision or adoption challenge. We do not arrive with a predetermined technology stack or expand the engagement into unrelated management consulting.
Engineering transformation
How do we introduce AI-assisted software, silicon, hardware, test, or product-development workflows without compromising IP, quality, or roadmap commitments?
Workflow assessment · architecture · tool evaluation · pilot-to-production
Enterprise knowledge
How do we make decades of documents, code, expertise, and institutional knowledge accessible to AI while preserving permissions and confidentiality?
Knowledge architecture · retrieval strategy · access controls · evaluation
Platform vs. best-of-breed
Should we standardize on an incumbent enterprise platform, adopt specialist AI tools, build selected capabilities internally, or combine them?
Requirements · RFP · vendor assessment · economic and risk tradeoffs
Agentic workflows
Where can agents safely coordinate work across systems, and where must a human remain in the loop?
Workflow design · orchestration · controls · acceptance criteria
Data, IP & sovereignty
What can leave the enterprise boundary? Which models may access sensitive information? What changes when teams and data span geographies?
Boundary architecture · permissions · vendor requirements · governance
AI governance
How do we let teams move quickly while managing security, privacy, legal, compliance, provenance, and model risk?
Policy translated into operating controls · ownership · monitoring
AI-native delivery
Clients should pay for judgment—not avoidable effort.
We are explicit about using AI in our own work. Research, synthesis, analysis, documentation, presentations, scenario development, and prototyping can all be accelerated. That allows a small group of experienced operators to do more work, faster, at lower total engagement cost.
AI accelerates. Senior Operators remain accountable.
- AI expands the amount of evidence we can examine
- AI shortens drafting and iteration cycles
- AI makes scenario testing and synthesis faster
- Senior Operators frame the problem and challenge the output
- Senior Operators own the recommendation and its consequences
Why the model is different
Leverage through AI—not layers of junior staff.
Traditional leverage model
Senior people sell the work
Junior teams perform much of the work
Hours and utilization drive economics
Strategy may be handed off at implementation
AI is added to the delivery model
→
DLG operating model
Senior Operators sell and deliver
AI absorbs low-value analytical and production effort
Clients pay for judgment and outcomes
Continuity through implementation and governance
AI is native to the delivery model
A firm of Senior Operators
Operating experience—not advisory résumés.
Three Senior Operators, over 100 years combined leading technology and engineering organizations—with particular depth in telecom and cable. Our advantage is not simply aggregate years of experience. It is the combination of having led those organizations, implemented change inside them, and now applying that judgment directly to AI transformation.
Joe Chow
Senior OperatorCommercial, product & operating leadership
44 years leading technology businesses through product, market, and operating transitions.
• President & GM, CommScope Home Networks — approximately $2B P&L and 1,500 employees• Built Cisco IPTV from zero to approximately $600M• Led Cisco Connected Devices business of approximately $2.7B• Helped scale Quantenna through record growth before its $1.1B acquisition
Gary Jamieson
Senior OperatorEngineering execution & scaled adoption
26 years leading engineering organizations, now personally building and shipping AI products.
• Former VP of Product Development at Comcast — led a 3,000+ engineer organization and $50M+ budget behind $80B+ in annual recurring revenue• Drove AI adoption at scale: virtual assistants, duplicate-detection, and productivity tools with 20%+ measured gains• As CEO of Vypian LLC, personally architected and shipped a production multi-tenant AI SaaS platform with a vendor-neutral, multi-LLM abstraction layer• Former Cisco software engineering leader; PhD in DSP and MBA
Khurram Qureshi
Senior OperatorApplied AI, ML & AI-native products
34 years leading global engineering organizations across cable, telecom, and mobile — now building AI products himself.
• Former VP of Software Engineering at CommScope — 150+ concurrent programs across Tier 1–3 operators• Former Executive Director at Comcast — led RDK-V/RDK-B platforms, shipped 2.75M+ next-gen video gateways• Founder & CEO of SelfAudit AI, a multi-agent LLM compliance platform with paying customers in regulated industries• Founded and scaled Fermat Software from zero to multi-million-dollar revenue; MS Data Science (UC Berkeley), MBA & BSECE (UT Austin)
Starting an engagement
Start bounded. Create decision clarity. Then decide what comes next.
The first engagement is deliberately scoped around a real decision, workflow, architecture question, or adoption problem. It does not require an open-ended transformation program.
Objectives
Owners, constraints, business outcomes, and decisions required.
Current state
Workflows, data, systems, tools, pilots, and readiness.
Value & risk
Prioritize opportunities and identify constraints.
Target state
Boundary, platforms, integration, and controls.
Next decisions
Priorities, pilots, implementation, governance, and metrics.
The output is decision clarity.
A fact-based plan showing where to focus, what environment is required, what decisions must be made, what should be piloted, and how success will be measured.
Crossing the Line · Insights
AI transformation is an operating discipline.
Our point of view focuses on the decisions that connect technical capability to production adoption.
01
Why AI Transformations Stall
Five predictable reasons promising initiatives fail to become durable operating capability.
02
Architecture Before Vendor Selection
Why data, IP boundaries, orchestration, controls, and workflow requirements should precede the RFP.
03
From Pilot to Adoption
The acceptance criteria, parallel operation, metrics, and human behavior required to earn production trust.
Ready to cross the line?
Bring us the AI decision that matters.
Whether the question is where to begin, how to architect the environment, which vendor to select, how to move a pilot into production, or how to govern adoption—we can start there.