Problem
All decade-scale public infrastructure programmes with many interdependent stakeholder streams enter an early Front-End Engineering and Design (FEED) phase. Business Cases are needed to be submitted for investment appraisal for sanction — typically both before and after. During this period:
- Contracts are not yet awarded, with many critical partners yet to be engaged.
- The “plan” in effect is a Gantt chart built in Primavera P6 — massive, granular, and largely speculative.
- Logic and scheduling dependencies lack realism (since so much is unknown), so the network has minimal variability or ability to model probabilistic logic.
- Non-planner stakeholders such as executives, commercial, operations — and even Project Managers — are exposed to unwieldy PDFs containing hundreds or thousands of schedule lines printed, then “reviewed” with highlighters, without real insight or trust.
This is a very common pathology in large programmes: early plans masquerade as schedules rather than strategic, flexible plans; and communication is trapped in technical artefacts few can consume or challenge meaningfully.
Challenge
Before intervention, programmes face two related problems:
1. Plan as schedule, not plan as strategy
Because of reliance on linear Gantt logic in P6, the schedule implicitly “locks in” decisions before stakeholder alignment, reducing strategic flexibility. Unknowns are suppressed rather than surfaced. Dependencies and inter-stream constraints are not visible or adjustable in a strategic view. Past decision events have proven that in a short time after sanction the stress on detailed schedules causes significant change and constant rework, resulting in the schedule looking completely different at the next gate or budget cycle.
2. Opaque communication and low stakeholder engagement
The only way non-planners can see or interact with a plan is via massive printed schedules or static PDFs. This means decision-makers are disconnected, unable to see interdependencies, trade-offs, risks, or the strategic narrative behind the timing decisions. The only option is to view the Plan on a Page or Level 0 plan — which conversely causes the opposite problem: not enough detail or information being communicated.
In short: planning artefacts are often too detailed too soon, opaque to most stakeholders, and unable to adapt or invite challenge.
Solution
Kaleido introduces a hybrid approach combining SharpCloud — a strategic portfolio visualisation and roadmapping tool — with AI integration to support conversational editing and governance.
1. Setting the narrative layer with SharpCloud
We treat the programme not as a single mega-schedule, but as a story with multiple chapters: design streams, transformation initiatives, risk exposure, delivery models, etc.
- Align project outcomes to goals and work backwards (right to left)
- Import schedule data via connectors (SharpCloud supports P6/Oracle connectors)
- Build “stories” — SharpCloud’s metaphor for narrative viewsets — that unify delivery, transformation, and risk streams
- Map dependencies and interfaces visually, allowing stakeholders to explore cause and effect, not just tasks
- Hold risk workshops and interface mapping inside SharpCloud: stakeholders see real-time visualisations of delay ripple effects or constraint bottlenecks
- Build different role views (executive, delivery, commercial) so each stakeholder sees only their relevant slice of the narrative, using the Role-Based Access feature
This produces a living, interactive roadmap rather than a monolithic schedule.
2. Embedding AI via MCP (Model Context Protocol) for conversational editing
To streamline change control, consistency, and speed, the Kaleido team worked with the SharpCloud technical team to integrate a Model Context Protocol (MCP) server — enabling a conversational/agent interface layered on top of the SharpCloud story model:
- Users make comments or change requests in “collaboration” mode on stories
- Change requests can be expressed in plain language (e.g. “slide design stream 12 months later”, “add dependency from X to Y”, “remove buffer here”)
- The AI agent, via MCP, interprets the request, validates it against the story data model, and applies the change to the underlying SharpCloud story model
- The creator sees the change materialise; no manual rework or re-editing of multiple visualisations is needed
- New stories can be generated from a blank slate by prompting the AI with high-level structure and vision
- Template story frameworks are built; the AI instantiates and adapts them based on context
Configuration control is preserved: only authorised “creator” roles can author stories; collaborators comment and suggest, but cannot freely override. The AI acts as the gatekeeper to enforce model integrity and consistency.
Partner: SharpCloud transforms traditional project portfolio management through sophisticated relationship mapping and visualisation technology. The platform excels at uncovering hidden interdependencies across complex portfolios, enabling dynamic “what-if” scenario planning that goes far beyond static Gantt charts. Their capability to connect projects to strategic goals, navigate complexity through relationship visualisation, and provide holistic knowledge integration makes them an essential component of our strategic roadmapping and capability mapping services.