Tag: contract-first

  • Contextual Scaffolding Framework

    Contextual Scaffolding

    a framework


    Transparency label: AI-assisted

    Creating a valuable knowledge-based product with AI depends on two things: the AI must clearly understand what’s required, and it must stay aligned with that requirement as the work progresses. Our solution is contextual scaffolding — a structured yet lightweight method that keeps the AI grounded and relevant from start to finish.

    A key feature of this approach is that the AI collaborates in building the scaffolding itself, starting with the initial programme of work. This ensures the AI is involved from the outset, shaping the context it will work within and adapting it as the project evolves.

    Contextual scaffolding works through two complementary structures. The phase model shapes the entire lifecycle, from initial concept to operational use. The project model organises collaborative work within each phase, ensuring the AI works with current priorities, materials, and deliverables.

    Both models are deliberately flexible, applied only as firmly as the situation demands, so that structure supports progress without stifling it.


    Phase Model (Cradle-to-Grave Structure)

    The phase model provides end-to-end structure across the full lifecycle of the product.

    1. Concept — Explore potential benefits, clarify purpose, and draft a business case.
    2. Definition — Specify the intended final deliverable in sufficient detail to support structured development.
    3. Development — Develop and test the final product.
    4. Acceptance — Validate the deliverable in the hands of its intended users, ensure it meets defined needs, and bring it into use.
    5. Operation and Maintenance — Use, monitor, and refine the product until it is retired.

    Not every phase will require a full project structure, but where structured collaboration is needed, the project model applies.


    Project Model (Within-Phase Scaffolding)

    Within each phase, a project-based approach provides structure for collaborative delivery.

    1. Defined Goal

    The project is established to deliver a clear, knowledge-based product. Typically, this will be the output required from a phase.

    2. Project-Based Structure

    Work within the phase is organised into discrete stages, each with defined objectives and outputs.

    3. Stage Gates

    Progression from one stage to the next requires a quality review of outputs. Only satisfactory products allow forward movement.

    4. Product-Focused

    Each stage produces one or more interim products. Most are “technical” in nature and build towards the final deliverable; some are to manage progress.

    5. Human-AI Collaboration

    The AI acts as a collaborator.

    • The human is the decision-maker.
    • The AI adapts to the nature of work at hand.
    • Because contextual scaffolding keeps pace with the progress of work, the AI remains grounded and its responses stay relevant and precise.

    6. Adapted PRINCE2 Principles

    Draws selectively from PRINCE2, including:

    • Product-based planning
    • Management by exception (e.g. gates)
    • Clear roles and responsibilities
    • Embedded quality control

    7. Quality Management

    A simplified quality management process ensures control:

    • Each product has a defined description or acceptance criteria
    • Quality is assessed at stage gates
    • Feedback loops enable iteration and correction

    Example: Preparing a Competitive Tender

    When preparing a competitive tender, the required knowledge product is a complete, high-quality bid submission. However, at the outset, the scope, fulfillment strategy, resource requirements, and pricing model will not be clear.

    Phase Model in Action
    The work progresses through phases – starting with a concept phase to assess feasibility and value, followed by definition (specifying scope and win themes), development (producing the actual bid), acceptance (review and sign-off), and finally operation and maintenance (post-submission follow-up or revision).

    Project Model in Each Phase
    Within each phase, a lightweight project structure is applied to guide collaboration with the AI. For example, in the definition phase, the AI helps analyse requirements, map obligations, and develop outline responses. In the development phase, it helps draft, refine, and format bid content.

    Contextual Scaffolding
    At every stage, contextual scaffolding ensures the AI is working with the right background, priorities, and current materials. Thus it can focus on what matters most, and contribute in ways that are coherent, precise, and aligned with both the tender requirements and internal strategy.

    Transparency label justification. This document was developed through structured collaboration between Alec Fearon and ChatGPT. Alec provided the core ideas, framing, and sequence. ChatGPT contributed to the organisation, refinement, and drafting of the text, under Alec’s direction. The content reflects a co-developed understanding, with human oversight and final decisions throughout.

  • An emerging discipline?

    Transparency label: AI-assisted

    In a recent post, I observed that LLMs are coming to be seen as like computer operating systems, with prompts being the new application programs and context the new user interface.

    Our precision content prompt pack is a good example of that thinking. The pack contains a set of prompts designed to take an unstructured document (notes of a meeting, perhaps) and apply structure (precision content) to the information it contains. We do this because AIs perform better if we give them structured information.

    We apply the prompts in sequence, checking the AI’s output at each step in the sequence. In effect, it is a program that we run on the AI. 

    Contract-first prompting takes the idea further. It formalises the interaction between human and AI into a negotiated agreement about purpose, scope, constraints, and deliverables before any output is generated. This ensures that both sides – human and AI – share the same understanding of the the work to be done. The agreement also contains compliance mechanisms (eg summarisation, clarifying loops, and self-testing) for quality control.

    These ideas reframe the interaction between human and AI: not as simply issuing prompts, but as engineering the conditions under which the AI will perform. Not just prompting, more like contextual systems engineering.

    Transparency label justification: This diary post was drafted by Alec, with ChatGPT used to suggest edits, refine wording, and test the logic of specific formulations. Alec initiated the framing, decided the sequence of ideas, and approved the final structure and terminology.