The FDE framework: from a vague productivity request to delivery
This article interprets and extends Lawted's “FDE nine-part framework” shared on Xiaohongshu. The original framework belongs to Lawted; my focus is why it matters in FDE work and how to turn it into actionable client assessment.
Companies often begin conversations with an FDE or AI consultant by saying, “We want AI productivity.”
That sounds like a requirement, but is not one yet. It may mean cost reduction, digital transformation, messy departmental processes or simply following the AI trend.
Before designing agents, knowledge bases or automation, ask what kind of client this is, whether its problem concerns strategy, workflows or ROI, whether it needs governance infrastructure, a working system or lightweight tools, and who will accept the result against which criteria.
Lawted's framework breaks a vague request into sales, product and engineering approaches for different clients.
The framework's core structure
The horizontal dimension has three capability tracks:
Sales -> Product -> Engineering
The vertical dimension has three client types:
Large state-owned enterprises pursuing strategic initiatives
Progressive businesses with roughly 200 employees
Businesses with roughly 30 employees seeking cost and efficiency gains
Their combinations create nine working areas. These are not nine mandatory steps. The same phrase, “AI productivity,” implies different entry points and acceptance standards for different organizations.
First tier: large state-owned enterprises with strategic goals
These projects often start with strategic approval and governance rather than a small tool. Questions include alignment with strategy, budget and approval pathways, reporting in terms the organization accepts, security, audit, permissions, compliance, human collaboration and evaluation.
Sales: strategic project approval
Understand stakeholders' intentions, approval and budget pathways, strategic goals and compliance reporting language. Progress depends on whether a project fits the organization's strategic narrative and approval system, not only technical performance.
Product: an enterprise business blueprint
The need extends beyond isolated features to cross-department workflows, capability maps, permissions, responsibilities, approval and audit points. An FDE must make complex roles and processes explicit.
Engineering: governance-oriented AI infrastructure
The focus is security, traceability, permission isolation, human–AI collaboration and evaluation rather than a quick demonstration. The outcome is:
A strategic initiative implemented safely, compliantly and with auditability.
Second tier: progressive businesses with roughly 200 employees
These businesses have organizational complexity and visible workflow problems. They may not need a large governance platform, but a few plugins or browser automations are no longer enough.
Key concerns are cross-department coordination, multiple workflows, system integration and real organizational AI adoption.
Sales: define the pilot together
Identify core intentions, choose a pilot department, build a phased roadmap and define success. Ask which department should start, which workflow is representative, how a successful pilot expands and what leaders and business owners consider success.
Product: orchestrate complex workflows
The problem is often coordination rather than one task. Identify five to ten key workflows, roles, data flows, dependencies and exceptions.
Information moving between sales, delivery, support and finance through copying, WeChat groups, spreadsheets and system notes may reveal useful AI or workflow opportunities.
Engineering: an agent-based working system
Delivery may involve agents or WorkBuddy, skills and knowledge bases, Feishu or ERP integration and workflow state management.
An agent should be embedded in work, not isolated as a chatbot:
- Receive inputs from a system.
- Apply business rules.
- Use knowledge and tools.
- Generate a suggestion or result.
- Route it for human review.
- Write back to the system or advance the next step.
The outcome is:
Working workflows and genuine organizational use of AI.
Third tier: roughly 30-person businesses seeking cost savings
These companies usually should not begin with complex platforms or long strategic programs. They want savings, fewer errors, quick payback and visible effects in one or two workflows. Start with ROI and core processes.
Sales: scope around ROI
Identify savings, focus on recurring pain, fix the scope and price, and show a payback path. Leaders want to know hours saved, errors reduced, whether hiring can be avoided, whether existing staff can handle more orders and how soon investment pays back—not an expansive transformation narrative.
Product: map the core workflows
Start with one to three processes such as order follow-up, contracts, reconciliation, quotations, support replies or project delivery materials. Define inputs, outputs and acceptance clearly.
A small complete workflow that works, can be accepted and saves time is more appropriate than a complex blueprint.
Engineering: lightweight tools
Plugins, web pages, lightweight ERP, Feishu bots and simple automation may fit. The priorities are:
Fast, reliable, usable
Avoid architectural complexity for technical completeness. Reliable reductions in hours and errors constitute useful delivery. The outcome is:
Clear cost and efficiency benefits, with fewer hours and errors.
What makes the framework valuable
Its value is not merely listing nine boxes. It reminds FDEs of three things.
Do not treat every client as the same type
Large organizations emphasize strategy, governance, compliance, audit and coordination. Medium-sized businesses emphasize working processes, integration and adoption. Small businesses emphasize ROI, payback and errors. One pitch and solution for everyone easily misses the need.
FDE is not a single engineering role
Sales, product and engineering are all necessary. FDEs are neither purely salespeople, product managers nor engineers. They connect why the client invests and how decisions happen with workflow, roles, permissions, inputs, outputs, acceptance and technical integration.
Deliver business outcomes rather than AI features
For large state-owned enterprises, the outcome is strategic implementation with security, compliance and auditability. For a roughly 200-person business, it is working workflows and AI adoption. For a roughly 30-person business, it is measurable savings in time and errors.
The deliverable is a result the client recognizes, not simply an AI tool.
Avoid using headcount as the only criterion
Headcount is a starting point, not a rigid classification. Also consider knowledge-worker numbers, workflow complexity, system maturity, budget, leadership goals, security and compliance needs, and an identified pilot department.
A 30-person company may have sophisticated processes; a 200-person company may be loosely organized; a large enterprise may initially need only a local pilot. Use the framework to identify the business setting, then conduct interviews and workflow analysis.
Implications for FDE Skills
This fits particularly well in `fde-intake`, which should assess the setting, choose a strategic, workflow or ROI entry point, identify questions for the leader, decide whether further FDE investment is justified and select the next stage skill.
It can support early assessment:
Large state-owned enterprise -> strategy, governance, compliance, business blueprint
Roughly 200-person business -> pilot department, complex workflows, system integration
Roughly 30-person business -> ROI, savings, core workflows, lightweight tools
Then continue through:
fde-intake
-> fde-discovery-interview
-> fde-workflow-mapping
-> fde-pain-analysis
-> fde-ai-opportunity-assessment
-> fde-mvp-scoping
-> fde-solution-brief
My conclusion
Lawted's framework places FDE at the intersection of commercial judgement, product design and engineering, rather than reducing the role to building AI tools.
“We want AI productivity” is the start. The work is identifying the client, choosing the entry point, designing a deliverable path and producing business outcomes the client accepts.
The first step is not building AI. It is deciding how this enterprise can turn AI into business results.
Thanks for reading.