BOB Community Call

Forvis Mazars Data & AI Program  ·  June 2026

Growing the BOB community across Forvis Mazars

01 Agenda

Community Call · June 2026

Today's agenda

01

Platform Updates

Migration from Gaïa to Claude, dynamic model routing, and data confidentiality architecture.

02

Market Intelligence

Enriched data analysis and the new automated PowerPoint generation capability.

03

Data & Knowledge

Sector data collection progress, data molds, and alignment with the Data Office initiative.

04

Community

Internationalisation, AI training, staffing updates, and open discussion.

Format: top-down presentation of updates and context, followed by structured open exchange. Selected contributors will be invited to share feedback directly. Questions welcome throughout.

Program status · June 2026

BOB Next Gen Proposal program board

Backlog · 127
#501 P1Conversation history awarenessBOB should remember context across sessions #526 P1 · Market IntelGroupe Premium: full intelligence packValuation, competitive position, news #541 · Market IntelAdd executive summary slideOpening slide missing from MI decks #572 · Market IntelNo iteration on same promptRe-submitting returns identical deck #574 · Market IntelLanguage selection for FR companiesFrench companies should default to French
Todo · 11
#640 · Conseil AssuranceBOB regulatory watch moduleNew module for insurance sector #613 · BPCEAudio playback on proposal siteAdd "Ecoute-moi" buttons #602 · KlepierrePrepare CSRD benchmarkESG comparative analysis #586 · SanofiAI Assurance and Semantic GovernanceNew website section design #650 · SanofiFix assets for interactive mapMap images not loading
In Progress · 5
#641 P1 · Total EnergiesRFP: Mobilite InternationaleFull proposal response in progress #651 · SanofiBlaise excel integrationIntegrating updated financial model #580 · Trusted IntelligenceData services credential workflowNew module under development #581 · Trusted IntelligenceCV bank for slide training setAdding data services CVs
Review · 5
PR #654 · SanofiBlaise excel modificationsOpen PR awaiting review PR #649 · KlepierreAudit proposal documentationOpen PR awaiting review PR #648 · AO ProposalComprehensive analysis and benchmarkOpen PR awaiting review #653 · PlatformTest Claude API FMZIn review #656 · PlatformTechnical Reference Doc for Group ITIn review
Done · 100+
#621 · BPCECeline/Xavier call feedback appliedHero, team roles, tools, login #638 · Market IntelBug detection agent workflowImproved support and auto-detection #615 · BPCEProposal download wired to R2 PDFStorage on Cloudflare R2 #608 · KlepierrePeople Intelligence LinkedIn workflowConnection mapping delivered #637 · PlatformFix FM URL for AI model accessUpdated endpoint across all modules #582 · Market IntelLogo management and CORS helperShared infra improvements #583 · Platform bugMissing BOB login for Anne-SophieAccess config resolved #657 · PlatformGroup-level approval to deploy BOBValidated by Group AI governance

Platform Analytics

Strong engagement from active users, with significant headroom to grow

Platform usage analytics

Next Gen Proposal workspace · Last 90 days

  • Active users: 24 of 80 total
  • Sessions: 93
  • Sessions per user: 3.9
  • Avg. session duration: 6m 17s
  • Returning users: 15

Activation rate is 30% (24 of 80 users). Retention among active users is 63% (15 of 24 returning). The 3.9 sessions per user and 6-minute average session are strong engagement signals from those who have adopted the platform. The spike in early June aligns with the BPCE AO and Klepierre deliveries, followed by sustained activity through the month.

70% of users have not yet activated. Group deployment and the Microsoft Tenant rollout will bring the platform to all members.

02 Platform Updates

Platform Updates

Data sensitivity drives model and infrastructure routing

One API endpoint for users and agents. BOB classifies every request and routes it to the right model tier. The decision is invisible to the caller.

Incoming request User, agent, or app via unified API Sensitivity classifier Tags data: public, internal, confidential, restricted LiteLLM routing layer GAIA Internal Forvis Mazars AI Models Client files, engagement data restricted & confidential workloads stay on-premise No data leaves the Mazars perimeter Claude API Market data, sector intelligence public company information non-client outputs Public & internal data Response returned to caller confidential public Audit trail: every call logged with tier used, sensitivity tag, latency, and caller identity

How to read it

  • Every BOB request passes through a sensitivity classifier before reaching any model.
  • Confidential and restricted data : client files, engagement data : routes exclusively to GAIA, the internal Forvis Mazars AI model infrastructure. Nothing leaves the Mazars perimeter.
  • Public and internal data : market intelligence, sector analysis, non-client outputs : routes to the Claude API for best performance.
  • Users and agents call one endpoint. The routing decision is invisible.

Data security and model performance are not a trade-off. They are a routing decision.

Platform Updates

Integrating private Forvis Mazars data into the intelligence process

What makes BOB outputs different from anything Claude alone can produce: the private FM knowledge that never leaves our infrastructure.

Claude API

Publicly available data

  • Market data, sector intelligence, industry trends
  • Public company filings, annual reports, regulatory documents
  • News, analyst reports, competitive landscape
  • General reasoning, synthesis, and generation

Fast, high-quality, cost-efficient. Available to any model with internet access.

GAIA: Internal FM Models

Private Forvis Mazars knowledge

  • FM credentials and track record across engagements
  • Named people: partners, sector specialists, relationship history
  • Sector perspectives built from decades of client work
  • Proprietary competitive positioning and offering data
  • Client files, engagement data, internal documents

This is what external models cannot replicate. It stays on Mazars infrastructure.

The intelligence BOB produces is the combination of both. Public data gives context. Private FM data gives the firm's unique angle, relationships, and credentials. Neither alone produces the output. That combination is the product.

03 Market Intelligence

Market Intelligence

Richer company profiles from URD extraction

What is new

Enriched data layer

The Market Intelligence report structure is being extended beyond the current baseline. New data dimensions are extracted automatically from URDs (universal registration documents) and equivalent regulatory filings.

Added dimensions

Shareholders, subsidiaries & geography

  • Shareholder structure and ownership breakdown
  • Subsidiary mapping and group architecture
  • Geographic distribution of operations and revenue
  • Finer extraction from regulatory filings
Why it matters

Context Forvis Mazars can act on

The data that differentiates us is what external models cannot produce: our own credentials, relationship history, and sector perspective built from decades of client engagements.

Shareholder and subsidiary data unlocks cross-entity engagement opportunities and benchmarking that clients cannot easily do themselves.

Geographic distribution maps directly to our cross-border network : 40,000+ professionals across 100+ countries.

Market Intelligence

Automated presentation generation in Forvis Mazars format

Capability

From prompt to branded deck

A new HTML-based approach to slide generation: describe what you need, BOB produces a fully designed, Forvis Mazars-branded deck you can export to PDF or PPTX directly from the browser.

Why it matters

Breaking out of the PPTX format lock

  • PPTX is a closed binary format, hard to generate reliably with AI
  • HTML gives full control over layout, branding, and content
  • Decks are promptable: iterate and refine in plain language
  • One-click export to PDF or PPTX when ready to share
Deployment plan

V0 · V1 rollout

V0: dedicated agent for slide generation, completing the Market Intelligence module : end of month target.

V1: extended typologies, richer template library, direct export from the BOB interface.

Status

In active development

The slide generation agent is being built now as a complement to the Market Intelligence module. First version targeted before end of June 2026.

04 Data & Knowledge

Data & Knowledge

Sector data collection and knowledge structuring

Progress

Sector coverage: 2 of 9 received

Life Sciences and Financial Services are the first complete datasets, used as the template for structuring contributions from other practice leads. The remaining seven sectors are pending input from their respective leads.

Sector scope
✓ Life Sciences ✓ Financial Services ○ Healthcare ○ Public & Social Sector ○ Private Equity ○ Consumer ○ Manufacturing ○ Technology, Media & Telco ○ Energy & Infrastructure
Methodology

Adapting data molds per sector

The initial data collection mold is being refined. It was designed as a general-purpose structure but some sectors require a different schema.

The key principle: isolate what is unique to Forvis Mazars and cannot be retrieved from external models.

  • Forvis Mazars credentials and track record
  • Named people and internal relationship networks
  • Sector perspectives built from engagement history
  • Competitive positioning that is proprietary to the firm

Data & Knowledge

BOB and the Data Office initiative

Context

Forvis Mazars France Data Office project

The Data Office initiative, led by Sylvain, is formalising data governance across the firm: each data domain has an identified owner responsible for data quality, completeness, and maintenance.

BOB and the Data Office are converging on the same objective : structured, governed, reusable firm knowledge.

Alignment

Data owners per BOB module

  • RSE module: owner to be reconfirmed
  • Market Intelligence: owner to be reconfirmed
  • Tech: owner to be reconfirmed
  • Life Science: established, used as reference

Recurring checkpoints will be established with each data owner to maintain alignment and accelerate knowledge collection across modules.

Next step

Joint session with the Data Office

A joint working session with the Data Office is planned to align on knowledge governance and data ownership across BOB modules. Scope: domain assignments, collection roadmap, and shared standards for H2 2026.

Group Initiatives

BOB and the Proposal Factory

Context

A Group-level programme to industrialise proposals

The Proposal Factory aims to standardise and accelerate the production of client proposals across Forvis Mazars. BOB is a natural contributor to this programme, covering the intelligence and content layers.

BOB can serve as the backbone for organising internal and external knowledge through a graph-based approach: connecting people, sectors, clients, and insights into a structured, queryable knowledge layer that powers every step of the proposal workflow.

What BOB covers
  • Integration layer: connecting sources, tools, and workflows into a single interface
  • LLM orchestration: routing tasks to Claude API, GAIA, or other models based on data sensitivity
  • Content generation: drafting, structuring, and formatting outputs using the most appropriate model
  • Branded delivery: FM-formatted slide decks and documents, exportable to PDF or PPTX
Outside BOB scope

Where external tools remain necessary

  • Licensed market data: calls with providers such as AlphaSense, PitchBook, or Bloomberg to source proprietary market data
  • Client-specific context requiring partner or account team input
  • Final review, pricing, and partner sign-off
Next step

Mapping BOB into the Proposal Factory workflow

A working session is planned to position BOB modules within the Group Proposal Factory process: identify handoff points, define what BOB automates, and where human or licensed-data steps remain essential.

05 Community

Community

BOB cleared for deployment across the full Forvis Mazars group

Group AI validation

Jeroen Hofland has confirmed group-level approval

BOB has received formal AI validation from Jeroen Hofland to proceed at group scale. This is not a pilot extension. It is a confirmed mandate to deploy BOB as a group-wide programme across the full Forvis Mazars organisation.

Next step

Group Microsoft Tenant, powered by the CoE

Ali-Sultan Kirgizbaev and his Centre of Excellence team will lead deployment into the Group Microsoft Tenant. BOB moves from a private, purpose-built infrastructure to group-vetted, centrally governed Microsoft infrastructure.

What this milestone means

A programme recognised at the Group level

Group validation through the CoE signals a structural shift in how BOB operates and is supported.

  • Group-level AI sign-off confirmed by Jeroen Hofland
  • CoE led by Ali-Sultan Kirgizbaev provides deployment authority
  • Group Microsoft Tenant: shared, governed, scalable infrastructure
  • All member firms gain access under a unified programme

This is the moment the programme moves from proving value to delivering it at scale.

Community

AI training and team organisation

Training

AI acculturation for partners

AI tool adoption across the partnership is growing. Best practices for BOB and Copilot are being shared across teams, and a structured acculturation programme is in preparation to support the group rollout at scale.

Staffing

2 to 3 interns in rotation

Intern recruitment now flows through the Data Services pool. The objective is to maintain 2 to 3 interns on the project in rotation, providing continuity across cycles.

Resourcing

Dedicated permanent role planned

A dedicated permanent role is being established to strengthen programme delivery alongside the existing team. Confirmation is expected in Q3 2026.

All presentation materials from Community Calls and related sessions will be centralised in a shared knowledge folder and made available to all programme participants.

Follow-up

Programme commitments coming out of this call

Knowledge and community

  • Community Call materials and session recordings centralised in a shared folder, accessible to all programme participants
  • Sector lead re-engagement underway across RSE, Market, Tech, and adjacent practice areas to accelerate knowledge base coverage
  • Practice leads invited to present at upcoming calls to share experience and drive peer adoption

Data and governance

  • Data ownership confirmed per BOB module, with recurring checkpoints established for each domain
  • Joint working session with the Data Office to align on knowledge governance and the H2 2026 collection roadmap

Deployment and resourcing

  • Resourcing model defined to support France, UK, Netherlands, and Germany as the programme expands across member firms
  • Group Microsoft Tenant deployment pathway confirmed in collaboration with the Centre of Excellence
Next call

September 2026

Date to be confirmed. Submit agenda items on GitHub or via Teams before end of August.

A dedicated Teams group for BOB support and community discussion is being considered. Watch this space.

Thank you

Forvis Mazars Data & AI Program

data-ai@forvismazars.com  ·  forvismazars.com