Chapter 05

AI for
Advisory
Services

From fragmented experiments to repeatable advisory value. How Arup translates AI capability into credible, governed advisory propositions.

What the evidence shows

Multiple internal artefacts already exist — business case, briefing note, benchmarking, and delivery examples across UKIMEA and Americas.

What this means

Move from AI as internal productivity to AI as decision-grade advisory capability — with repeatable offers, methods, and governance.

Context

Why This Matters Now

The question is no longer "should advisory use AI?" — it's "which AI-enabled services can Arup credibly take to market first?"

Client demand is changing

Clients are increasingly asking how AI can improve delivery, decisions and outcomes — not just whether we have AI capability.

We already have practical proof

Arup teams are already applying AI on live projects, so we can build from real experience and credible examples.

Trust is our advantage

Our strength is responsible AI — with safety, ethics, auditability and human oversight built in. That is what gives clients confidence.

Portfolio View

Current Initiative Landscape

Initiatives cluster into four practical themes that underpin a coherent AI-in-Advisory playbook.

Theme 01

Operational AI & Data Productivity

Embedded Copilot use across day-to-day work — Outlook, Word, Excel, Teams — to lift baseline productivity.

CopilotM365Daily flow

Theme 02

Domain Agents & Decision Support

Project-scoped agents that act on curated data — risk, planning, commercial, BIM, IM. Human approves before write-back.

Project agentsSharePointAudit trail

Theme 03

Knowledge & Retrieval

Notebooks, Researcher, and curated agents that turn institutional knowledge into a queryable asset.

NotebooksResearcherCitations

Theme 04

Generative Content & Visualisation

From draft documents and decks to bespoke visuals and dashboards — quality at speed, brand-governed.

Word/PPTPower BIBrand kits

Evidence Base

Arup-Wide AI Case Studies

Real examples showing how AI strengthens advisory when it augments judgement and systematises expertise.

Decision Intelligence

Climate option appraisal

AI-assisted comparative analysis across decarbonisation pathways for a major UK infrastructure programme.

Cross-portfolio · cited in business case

Domain Agent

Multi-agent project intelligence

Five AI domain agents (risk, planning, commercial, BIM, IM) coordinated by an orchestrator. Discovery in flight; pilot targeted.

Mesut Pala · Darshan Ruikar

Knowledge & Retrieval

Project memory Notebook

Scoped Notebook holding project documents, recaps, and decisions — answers cited from corpus only.

UKIMEA Advisory · multiple projects

Generative Content

Custom Power BI risk visual

In-house visual built with Copilot + Claude in ~5 hours — no subscription cost, reusable across Arup.

Tapan Parmar

Decision Intelligence

National Highways change-control app

Copilot-driven planning and screen generation, replacing a multi-week manual design cycle. Reduced rework and scope changes.

Tim Williams

Domain Agent

BIMVeritas — Defra

Internally developed model validator. ISO 19650 / COBie compliance results in minutes, exported to Power BI.

Erkan Sahin

Operational AI

Copilot adoption sprint

Targeted enablement with prompt packs and Day workflow. Measured ~4 hours/week saved per advisor.

UKIMEA Advisory

Knowledge

Researcher pilot

Quota-disciplined deep research for proposals. Compresses 2-day landscape scans into ~1 hour with citations.

Pilot cohort · 2026 Q1

Domain Agent

Project briefing agent

Generates a tailored 1-pager for any project from SharePoint sources. Zero training time for new joiners.

UKIMEA · cross-project

Visualisation

Brand-kit governed visuals

AI image generation gated by Green/Amber/Red governance grid. Brand-approved swap-out before client send.

Comms & Advisory

What We Learn

Five Takeaways for Advisory

Patterns observed across Arup's AI projects — and what each means for how advisory teams work.

AI augments judgement

Outputs are inputs to a decision — never the decision itself.

Repeatable agents win

One-off prompts don't scale. Patterns and agents do.

Data trust is the gate

If the source data isn't curated, the AI output isn't usable.

Governance is a feature

Audit trail, citations, and approval flows are what clients buy.

Domain expertise is the moat

Advisory judgement is the differentiator AI can't replicate.

Advisory Service Lines

AI Across Arup's Three Advisory Service Lines

Arup Advisory is delivered through three official service lines — Business and Investor Advisory (BIA), Management Consulting (MC), and Programmes & Project Management (P&PM). Each line gets a different AI accelerator, mapped here to live internal evidence.

Service Line 01 · BIA

Business and Investor Advisory

AI-augmented option appraisal, transaction support, ESG & policy narrative, and evidence synthesis for investment-grade decisions.

Where AI helps

Comparative option appraisal with cited sources
Risk cluster analysis from a curated project corpus
Regulatory & ESG framing for board / investment committee papers

Typical buyer: Chief Financial Officer (CFO) · Investment Committee · Asset Owner

See evidence — Multi-Agent Project Intelligence

Service Line 02 · MC

Management Consulting

AI-augmented strategy, organisational diagnostics, and assets & operations analysis — domain agents working on live client data.

Where AI helps

Strategy synthesis from interviews, surveys, and benchmarks
Operating-model and people-data analytics
KPI / change-control consulting tooling (e.g. MetricHub)

Typical buyer: Director · Programme Director · Strategy Lead

See evidence — MetricHub (National Highways)

Service Line 03 · P&PM

Programmes & Project Management

AI-assisted programme assurance, risk surfacing, lessons retrieval, and document automation across complex capital programmes.

Where AI helps

Model assurance / QA-QC at scale (e.g. BIMVeritas)
Programme & project risk reporting (e.g. Risk Matrix Power BI Visual)
Lessons-learned retrieval across past project corpora

Typical buyer: Programme Director · Project Sponsor · Project Director

See evidence — BIMVeritas (Defra)

The Advisory Shift

What Changes When AI is Introduced

AI doesn't replace advisory professionals. It changes the ratio of time spent on manual synthesis versus high-value judgement.

Time was spent here

Hand-synthesising source material
Re-formatting reports for each audience
Stitching status from disparate sources
Producing rather than challenging

Time now goes here

Framing the right question
Stress-testing AI outputs
Client-facing judgement calls
Designing repeatable agent patterns

Walk this shift up close in The Users Day →

Adoption Path

Four Implementation Levels

Start individual. Expand to team, then project, then enterprise — once governance is in place.

Individual

Daily Copilot use across Outlook, Word, Excel

Team

Shared prompts, agents, and workflow patterns

Project

Project-scoped agents on curated SharePoint data

Enterprise

Governed, audited agents in client-facing work

What to Build

The Agent Maturity Ladder

Start retrieval-first. Earn the right to task agents. Approach autonomy only when the audit trail is solid.

Today

Retrieval Agents

Does Answers questions from a curated corpus with citations.

Needs Source curation. Scoped grounding. Notebook discipline.

Arup is Pilot & scale across UKIMEA Advisory.

Next 12 months

Task Agents

Does Performs domain-specific tasks — risk proposal, schedule check, RFI triage.

Needs Approval flows. Write-back governance. Audit trail.

Arup is Multi-agent project intelligence in discovery; pilot targeted.

Horizon

Autonomous Agents

Does Acts unattended within bounded mandates — orchestration across agents.

Needs Mature governance, contractual frameworks, tested rollback.

Arup is Earning the right — not deployed yet.

Start governed-retrieval. Earn the right to task agents. Approach autonomy only when the audit trail is solid.

Ready to Use

Advisory Prompt Packs

Copy-paste prompts for the four highest-leverage advisory tasks.

Briefing & narrative · Executive 1-pager

Role: You are a senior Arup advisor preparing material for a client C-suite. Context: The brief covers [topic]. Audience is [CEO / CFO / board], who care most about [risk / cost / strategic fit]. Task: Distil the brief into a 1-page executive note. Surface the decision being asked, three options, recommended option with rationale, and 30/60/90-day actions. Format: 1 page. Headline · 3 bullets per option · recommendation in bold · timeline as a small table. Guardrails: No technical jargon. Flag uncertainty explicitly. Do not invent figures.

Risk clustering · From a long register

Role: You are a senior consultant analysing a project risk register. Context: The attached register has [N] risks across [project]. We need a board-ready view, not a long list. Task: Cluster risks into 4–6 themes. For each cluster: theme name, top 3 risks (verbatim), aggregate likelihood/impact, the single most important mitigation owner. Format: Table. Themes ordered by aggregate exposure, highest first. Guardrails: Use only the register text. Don't merge risks that look similar but have different owners. Flag any ambiguous wording.

Climate framing · Decarbonisation narrative

Role: You are a senior climate-and-infrastructure advisor. Context: The client is [sector]. They need a board narrative for their decarbonisation pathway covering [timeframe], constrained by [regulation / market / capital]. Task: Draft a 1-page narrative — current trajectory, three pathway options with carbon & cost trade-off, recommended path, and the regulatory/policy assumptions it depends on. Format: 4 short sections. Trade-off as a small comparison table. Guardrails: Cite sources for every figure. Don't conflate net-zero claims with science-based targets. Flag where the policy environment is unsettled.

Stakeholder positioning · Map & messaging

Role: You are a senior consultant shaping engagement for a complex programme. Context: The programme is [name]. Stakeholders include [list]. Recent positions captured in [Notebook / consultation transcript]. Task: Build a stakeholder map — for each stakeholder: current position, primary concern, the single message that lands with them, and the next engagement action. Format: Table, one row per stakeholder. Sort by influence × current resistance. Guardrails: Use only positions captured in the source material. Mark inferred positions as "inferred" — do not present as fact.

Route to Market

From Initiatives to Client Offers

A practical pathway to translate current material into propositions leadership can pilot, refine, and scale.

Initiative

Internal experiments — proving capability and pattern.

To advance Internal evidence, named owner.

Arup Multiple — landscape covered.

Pilot

Live deployment on a real project — one team, one client, scoped.

To advance Outcomes measured. Governance signed off.

Arup Several in flight.

Productised Offer

Repeatable proposition with named deliverables, pricing, and a sell sheet.

To advance Two paying clients. Margin proven.

Arup Gap — the focus of the next 90 days.

Scaled Service

Established part of the advisory portfolio — multi-region, multi-team.

To advance Operating model, capability pipeline, P&L line.

Arup Horizon target.

Next Steps

90-Day Action Plan

Move quickly without over-scaling too early — consolidate, pilot, learn, formalise.

Days 0–30

Consolidate

Audit existing initiatives across themes
Confirm the priority service lines to pursue (BIA · MC · P&PM)
Stand up a single playbook owner
Confirm trust baseline applied across teams

Days 30–60

Pilot

Run one client pilot per priority service line
Define outcome metrics before kick-off
Capture every artefact in a shared Notebook
Brief leadership weekly with verbatim quotes

Days 60–90

Formalise

Convert each pilot into a productised offer
Publish sell-sheet with deliverables & pricing
Brief client leads & portfolio teams
Lock the operating model for the next quarter

What Must Be in Place

Enablers & Guardrails

Enablers

Curated data

SharePoint, Notebooks, and Loop scoped per project — no curation, no AI value.

Governance

Approval flows, audit trails, and EDP-by-default across every advisory engagement.

Capability

Every advisor fluent in the trust baseline and the 5-part prompt template.

Client trust

Transparent transcription, citations on every claim, and a clear "AI-assisted" disclosure.

Guardrails

Trust baseline

EDP, scoped grounding, transcription consent, and quota discipline — applied to every workflow.

IP & confidentiality

Client material in scoped Notebooks. Never in free-tier Chat. Disclosure agreed up front.

Model selection

Match model to task. Frontier models for reasoning, not for low-stakes drafts.

Audit trail

Every AI-generated artefact carries the prompt, the model, and the source set used.

Winning Position

Use AI to enhance judgement, structure evidence, and accelerate delivery — with governance as the moat.

Domain expertise + AI capability + Governance

Immediate ask: consolidate the current asset base and stand up a small number of well-governed advisory offers.

AI forAdvisoryServices

Why This Matters Now

Client demand is changing

We already have practical proof

Trust is our advantage

Current Initiative Landscape

Operational AI & Data Productivity

Domain Agents & Decision Support

Knowledge & Retrieval

Generative Content & Visualisation

Arup-Wide AI Case Studies

Climate option appraisal

Multi-agent project intelligence

Project memory Notebook

Custom Power BI risk visual

National Highways change-control app

BIMVeritas — Defra

Copilot adoption sprint

Researcher pilot

Project briefing agent

Brand-kit governed visuals

Five Takeaways for Advisory

AI augments judgement

Repeatable agents win

Data trust is the gate

Governance is a feature

Domain expertise is the moat

AI Across Arup's Three Advisory Service Lines

Business and Investor Advisory

Management Consulting

Programmes & Project Management

What Changes When AI is Introduced

Four Implementation Levels

Individual

Team

Project

Enterprise

The Agent Maturity Ladder

Retrieval Agents

Task Agents

Autonomous Agents

Advisory Prompt Packs

From Initiatives to Client Offers

Initiative

Pilot

Productised Offer

Scaled Service

90-Day Action Plan

Consolidate

Pilot

Formalise

Enablers & Guardrails

Enablers

Curated data

Governance

Capability

Client trust

Guardrails

Trust baseline

IP & confidentiality

Model selection

Audit trail

Where to Go from Here

The Users Day

AI in Action

Foundations & Tools

Use AI to enhance judgement, structure evidence, and accelerate delivery — with governance as the moat.

AI for
Advisory
Services