Why we call them signals, not cases — and why that distinction matters.
A case study is a marketing artefact. A signal is evidence.
The difference matters. Most consultancy case studies are written to sell — they start from the outcome, omit what didn't work, and polish the narrative until the causality looks inevitable. We don't find that useful, and we suspect you don't either. What we publish here instead are signals: structured accounts of Phase Three work, anonymised where the client has not explicitly given permission to be named, focused on what the engagement proved about Phase Three rather than what it proved about us.
Each signal below follows the same shape — industry, scale, the Phase Three question the business faced, the shape of the engagement, what shifted, and what it signals about Phase Three more broadly. We name clients only with explicit permission. We quote specific numbers only where those numbers are verifiable and permissioned. When neither is available, we describe the outcome qualitatively and let you judge from the structure.
The signals below are anonymised accounts of the team’s AI transformation work, published as of July 2026. We name clients only with explicit permission; where we don’t have it, we keep the outcome qualitative — sector, scale, what was built, and what shifted — and let the structure speak. More will follow.
A small sales function couldn't cover its outbound pipeline without headcount it couldn't justify. The team built a full outbound automation stack — CRM-integrated, human-in-the-loop, with automated lead sourcing and enrichment — and the same people took on materially more pipeline.
A luxury-brand marketing team needed to move faster on content while staying strictly inside brand guidelines. The team built a brand-safe content workflow on their existing tools and trained the group to run it — quicker production, brand compliance preserved, no custom-model risk introduced.
Growth depended on catching contract-renewal windows and qualifying leads faster than people could manage. The team built AI lead qualification, contract-window detection, automated outbound, and recycled-materials marketplace intelligence — framed for a CEO in margin and capacity terms.
A fragmented go-to-market and back office capped how fast new locations could ramp. The team delivered a four-system transformation across demand generation, qualification, and operational handoff — a repeatable motion designed to scale without proportional headcount.
The channel team needed to prioritise partners for an exclusivity push across a multi-state territory. The team built a two-level dashboard — a leadership rollup and a branch-level drill-down carrying a scored prospect list — so leaders see territory performance at a glance and act on a ranked target list.
Drivers defaulted to slower no-cost disposal sites over faster paid ones, and dispatch ran by hand. The team built dynamic route-and-site cost optimisation plus an automated dispatch layer — two clean, dollarisable anchors: recovered driver hours and reduced dependence on manual dispatch.
Complex rental pricing lived across fragmented systems and spreadsheets with no single source of truth. The team built an ontology layer to unify the data, then an AI-native pricing and decision dashboard on top — pricing grounded in a connected model rather than tribal knowledge.
Institutional knowledge and operating data were scattered across systems and people. The team built an enterprise ontology — a single connected knowledge base with source-grounded reasoning on top. This engagement became the internal benchmark for how the work is framed and measured.
A complex multi-state operation ran on disconnected systems that made enterprise-wide visibility difficult. The team designed an enterprise ontology layer spanning the footprint, with a prioritised set of use cases sequenced for phased delivery.
Valuable inference technology sat unused, with intellectual property lapsing. The team wrapped an LLM layer around the existing engine, revived and refiled the patent, and built a dual consumer-and-business health agent — turning dormant IP into a defensible product, with direct support to the company's capital raise.
A market opportunity existed with no product to capture it. The team built the entire thing from zero — a knowledge graph, a revenue model, an AI agent suite, and a B2B marketplace matching producers, processors, and buyers.
The business ran on generic software built for no one in particular. The team built a full agentic operating system — a CRM with embedded agents, a guest-and-VIP management agent, and an industry-specific knowledge graph.
The firm wanted an edge in sourcing distressed and turnaround deals, and a way to lift AI capability across its portfolio. The team built three workstreams — a white-label AI service line, contact intelligence for deal sourcing, and a portfolio-company transformation program.
At a century-old mutual carrier, content and campaign work moved slowly through multi-day approval chains, with no governed way to use AI safely. The team built governed AI assistants, prompt libraries, and a governance framework, applied to a live flagship campaign — approvals compressed to same-day, with controls in place rather than bypassed.
Inside a global technology company, AI tooling had sprawled to 245 uncontrolled tools with no governance or shared standard. The team delivered a sprawl discovery report, a literacy program for 150 employees, an experimentation framework, prompt libraries, and an internal champion program — sprawl replaced by something governed and teachable.
At an international law firm, attorneys and staff had no firm-wide standard or training for safe, useful AI. The team ran executive workshops and firm-wide education, rolled out an enterprise assistant, and led with AI note-takers as an immediate quick win — a governed baseline of fluency.
Evidence earns the claim. Claims without evidence erode it.
We are deliberate about what goes on this page and what doesn't. Engagements we are currently running may never appear here — either because the client prefers discretion, or because the engagement is ongoing and the outcome cannot be fairly characterised until later, or because what happened in the engagement is genuinely not suitable for public signal work. None of those absences should be read as a problem. Most of our most valuable work is not, and probably will not be, on this page.
What appears here is a deliberate subset — signals we have permission to share, that carry evidence, and that meaningfully illustrate something about Phase Three. The integrity of that filter is more important to us than the density of the grid.
Want to see a signal from your sector?
We can often share additional work under NDA in the enquiry conversation. Tell us the question you're trying to answer.
Enquire →