The Media Copilot

Orchestra

AI Advisory Engagement · January 2026

Orchestra AI
Advisory Portal

Your central resource for the AI advisory engagement with The Media Copilot — findings, strategy, and deliverables in one place.

8
Interviews
Stakeholder voices
117
Insights
Observations gathered
250+
Trained
People across the org

This portal captures the full arc of The Media Copilot's AI advisory work with Orchestra. Starting in late 2025, we conducted a structured discovery process — interviewing stakeholders across the organization, synthesizing what we heard into actionable recommendations, and delivering a roadmap that connects AI adoption to real business outcomes.

The work is organized around three pillars: making AI safe and consistent, building systems that compound over time ("build once, reuse often"), and directing those gains at client value. The deliverables in this portal bring those pillars to life.

"Orchestra has real AI momentum — the next phase is turning scattered wins into governed scale, reusable intelligence, and visible client value."

01
Safe & Consistent
Clear rules, guardrails, and governance that let teams move faster with confidence.
→ Work faster with confidence
02
Build Once, Reuse
Centralize knowledge and scale the best ideas instead of reinventing from scratch.
→ Stop reinventing, start compounding
03
Client Value
Aim every efficiency gain at client outcomes — faster delivery, better insights, new offerings.
→ Win more, keep clients longer
Advisory Presentation
Narrative-driven readback covering all findings, recommendations, and the strategic roadmap
AI Policy
Visual, plain-language AI usage policy for Orchestra — what's allowed, what needs approval, and who to ask

The full presentation walks through what we discovered, the tensions we identified, and a clear path forward across all three strategic pillars.

AI Advisory Presentation — January 2026

What We Discovered. Where We Go Next.

A narrative-driven readback designed to validate stakeholder input and present a clear path forward.

Orchestra has real AI momentum — the next phase is turning scattered wins into governed scale, reusable intelligence, and visible client value.
8
Interviews
Stakeholder voices
117
Insights
Observations gathered
250+
Trained
People across the org

Your Voices Shaped This

Across 8 stakeholder interviews, clear patterns emerged — both momentum and genuine friction.

Momentum
Real Momentum
Teams are already experimenting with AI across the organization.
ChatGPT, Otter, custom workflows — adoption is happening organically.
Momentum
Training Win
Outstanding attendance for fall AI training sessions.
Broader understanding of structured prompting, AI assistants, and deep research.
Momentum
Applying AI
ChatGPT of the Month program nurtures skills.
Incentivizes and rewards people for putting AI into practice and sharing what works.
Challenge
GEO Opportunity
"We need to know how to advise our clients on how to reach their audiences in an AI world."
Clients are asking how AI changes visibility and authority — Orchestra needs answers.
— Kate Riley
Challenge
Governance
"A single file can destroy a client relationship."
Competing clients, confidential data, porous containers — the risk is real and needs a formal response.
— Louis-Philippe Cavallo, CFO

The Challenge You Face

Client pressure on value and speed vs. fragmented internal workflows and knowledge.

Client Pressure
  • Faster turnaround
  • More for less
  • Differentiated value
  • Proactive insights
vs.
Internal Reality
  • Fragmented workflows
  • Scattered knowledge
  • Uneven adoption
  • Limited governance
The Opportunity
Move from reactive service to proactive, agentic operations

The Path Forward

Three pillars to anchor every decision.

01
Safe & Consistent
Make AI safe, easy, and consistent so teams can work faster with confidence.
→ Work faster with confidence
02
Build Once, Reuse
Centralize intelligence and scale the best ideas across the organization.
→ Stop reinventing, start compounding
03
Client Value
Aim every efficiency gain at client outcomes — faster, better, more proactive.
→ Win more work, keep clients longer
Act One
Safe and Consistent
Streamline AI so teams can work faster with confidence

Guardrails That Unlock Speed

Clear rules build the confidence to move fast without creating risk.

Data & Privacy
Clear rules on what can and cannot go into AI tools.
  • No client data in personal tools
  • Enterprise ChatGPT for sensitive work
  • Data classification before AI use
Client & Reputational Risk
Approval requirements for client-facing AI output.
  • Steering group reviews new use cases
  • Client-facing AI gets extra scrutiny
  • Clear escalation path
Human Review
Every external output reviewed by a human before it leaves the building.
  • No auto-send to clients
  • Editorial mindset required
  • AI drafts, humans decide

Two Lanes for Innovation

Protect experimentation while creating a path to scale.

Lane 1
Personal Sandbox
Experiment freely within guardrails
  • Custom GPTs for personal use
  • Approved tools only
  • No sensitive data
  • Human review before sharing
  • Share back: prompts & wins
Encourages experimentation
Lane 2
Shared & Client-Impacting
Formal path to scale
  • Shared across teams
  • Client-facing work
  • Integrated systems
Requires
  • Lightweight review by steering group
  • Standard pilot template
  • Clear success metrics
  • Scale or sunset decision
Creates path to scale best ideas
Q1 Quick Win
Shadow AI Amnesty
Surface hidden tools, reduce risk, and find the best ideas already happening inside the organization.
1

Capture

Survey what tools people are already using and why they chose them.

2

Migrate

Move identified needs into approved tools or formalize them as official pilots.

Act Two
Build Once, Reuse Often
Centralize intelligence to scale the best ideas across the organization

Three Pilots to Launch

Ready to launch in Q1. Scale or sunset after 90 days.

Click any pilot card to see the full brief, including before/after scenarios and success metrics.

Centralized Knowledge Hub

Who Uses It
All teams, especially new business and account leads
What It Replaces
Searching across folders, threads, old decks
Success Metric
Find relevant past work in under 30 seconds
Done Means
Scale across org or sunset after 90 days
Before
Search 5 different folders, Slack, email
After
One place to ask, retrieve context and assets
Click to expand ↓

Client Health Digest

Who Uses It
Account leads, CFO for at-risk monitoring
What It Replaces
Manual pulse checks, scattered notes
Success Metric
Flag at-risk clients before they churn
Done Means
Proven value on one complex account → roll out
Before
Account health lives in scattered notes, inboxes, and memory until something goes wrong.
After
A daily brief surfaces risks, opportunities, and next steps early enough to act.
Click to expand ↓

Proposal Assistant

Who Uses It
New business team
What It Replaces
Starting from scratch on every pitch
Success Metric
Reduce pitch prep from 1.5 hours to 30 minutes
Done Means
Active use by new business team
Before
Proposals start from scratch, with people chasing past language and reinventing structure under deadline.
After
A guided draft pulls from approved templates and past wins, so teams produce faster, more consistent proposals.
Click to expand ↓
Act Three
Client Value
Aim every gain at client outcomes — win more work and keep clients longer

Where Clients Feel the Difference

Better delivery today. New advisory offerings tomorrow.

Better Delivery
Improve current work
Faster turnaround on deliverables
Consistent quality across deliverables
Proactive insights from client data
New Advisory Offerings
Create new value
New Offering

AEO: Answer Engine Optimization

What It Is
Helping clients show up accurately and prominently in AI answers and summaries
Why It Matters
AI systems are becoming a first stop for information, affecting visibility, authority, and demand
What Orchestra Offers
Practical assessments and playbooks for content, PR, and measurement
Differentiator
First-mover advantage in an emerging space

Roadmap

Sequenced for steady adoption and scope control.

Now
0–3 months
  • Publish policy, train on it
  • Create steering group
  • Run shadow AI amnesty
  • Kick off AI pilots
  • Begin monthly trainings
Next
4–6 months
  • Continue monthly trainings
  • Client GPT rollout
  • Measure and iterate pilots
Later
6+ months
  • Unified AI environment
  • AI Lab for ongoing innovation
  • System-level integrations
  • Ongoing training

How We Track Progress

Four categories of measurable outcomes.

AI Adoption
Percentage of staff using approved AI tools at least weekly.
Target: Grow to a clear majority of staff, with usage distributed across functions.
Workflow Impact
Time reduction on benchmark tasks after AI support is introduced.
Target: Clear, measured reductions — e.g., 30–50% time savings on specific tasks.
Client Value
Client satisfaction scores and increase in new business wins.
Target: Higher scores that mention speed or quality; measurable new business growth.
Culture & Skills
Percentage of staff who feel confident using AI in their role.
Target: Clear upward trend over successive surveys, with low AI-related anxiety.

Ready to Move Forward

What we need from leadership to begin.

Crawl, Walk, Run
Crawl
Establish rules, foundations, and norms
Walk
Run pilots, measure, repeat what works
Run
Move from separate projects to integrated systems
Leadership Decisions Required
1
Name steering group membership
Leadership
2
Approve amnesty window
IT + Comms
3
Pick the first 3 pilots
COO + Practice Leads
4
Commit to training cadence
HR + Culture
5
Align on KPI reporting cadence
Leadership
Full Report & Source Materials
Orchestra AI Discovery Report
The complete report, raw findings, and supporting materials are available in the shared Google Drive folder.
Open in Google Drive
Orchestra AI Policy · 2026

How We Use AI

A practical guide for every Orchestra employee and contractor — what's allowed, what needs approval, and what's off-limits.

1
Use only Orchestra-approved tools for anything involving client or internal information.
2
AI drafts, humans decide. Every external output needs a human review before it leaves the building.
3
When in doubt, treat information as sensitive — and ask before you act.
How to read this
Allowed on approved tools
Requires review & approval
Never allowed

"Requires review and approval" means submitting a request through the AI intake form to the policy owner and AI council, who will approve, deny, or approve with guardrails.

Data & Privacy
Protecting client information and keeping data separated
+
Allowed
Drafting, summarizing, rewriting, and analysis using Orchestra-approved enterprise tools (e.g., ChatGPT Enterprise), with client or internal confidential info — keeping each client's information separate from other clients, sharing outputs only within the correct client workspace or team channel, and minimizing what you paste or upload to what is needed for the task
Creating internal checklists, outlines, and first-pass drafts — as long as a human reviews before any client delivery
Requires Review & Approval
Using any non-approved tool, personal account, browser extension, or unvetted AI service
Connecting AI tools to Orchestra systems (Google Drive, email, calendar, Slack, ticketing) or to client systems
Any plan to fine-tune, train, or otherwise adapt a model using Orchestra or client data
Use of personal information as AI input when it is sensitive, unnecessary for the task, or used at scale
Never Allowed
Mixing two clients' confidential information in the same workspace, thread, project, or dataset
Copying raw outputs into external-facing work without human review
Entering credentials, API keys, or access tokens into any AI tool
Inputting sensitive personal information (numerical identifiers, passwords, financials)
Using AI with client information
1
Classify the input as public, internal, client confidential, or personal data
2
If internal or client confidential, use only approved enterprise tools
3
Use the appropriate client workspace, project, or approved client GPT
4
Save outputs in the appropriate client workspace — do not combine confidential information from different clients in the same chat, project, GPT, or knowledge base
5
If the task requires broader data access, request IT review and approval before proceeding
If something goes wrong
1
Stop and preserve context: which tool, what was shared, when
2
Notify the IT and security owner and the policy owner
3
Follow internal incident steps and client contract requirements for client notification if required, in coordination with Legal
4
Record the incident for quarterly review and process improvement
Writing & Assistance
AI helps the work — people remain accountable for it
+
Allowed
First-pass drafting for emails, outlines, press materials, Q&A prep, summaries, and meeting materials
Rewriting for tone, clarity, structure, and length
Creating checklists, project plans, agendas, and internal templates
Building interview questions, internal briefing docs, media lists, and stakeholder messaging frameworks
Requires Review & Approval
Any content that makes product claims, particularly in regulated industries like finance and pharma — review with the client before use
Any content containing legal, compliance, regulatory, or contractual statements or claims
Any plan to publish or send AI-assisted copy externally without human review and edits
How AI-assisted drafting should work
1
Use the approved client workspace, project, or client GPT for any client-confidential material
2
Provide the AI with the source material, context, and constraints needed to do the task well
3
Verify facts, names, dates, claims, quotes, and citations against primary sources
4
Confirm any third-party material used as input is approved for the project and doesn't violate license, contract, copyright, or client restrictions
5
Review outputs for accuracy, tone, client appropriateness, and risks like bias or harmful stereotypes
6
Add a second reviewer for high-stakes deliverables
7
Save a brief record of what was AI-assisted if the project requires traceability
Transparency & Disclosure
Being honest about AI use — internally and with clients
+
Allowed
Being open internally about when AI was used for assistance during drafting and review
Using a simple internal note like "AI-assisted draft — reviewed and edited"
Requires Review & Approval
Any client request for a formal disclosure statement
Any public-facing disclosure language — website statements, pitch language, boilerplate
Situations where disclosure could affect trust, compliance, or contract terms
Never Allowed
Misrepresenting AI-generated or AI-assisted work as fully human-created when disclosure is required for public-facing deliverables
Using AI on a client account that has explicitly prohibited it — client contractual restrictions supersede this policy for all work on that account
Client account with an AI restriction
1
Record the restriction in the account's working procedures
2
Confirm the account lead and core team are aware of the restriction
3
Use a clear internal flag so teams avoid accidental use
4
If an exception is requested, route through the policy owner, account lead, and any required legal or risk reviewer
Intellectual Property & Copyright
Respecting rights — for text, images, and everything in between
+
Allowed
Summarizing or rewriting materials you have rights to use for the task
Editing Orchestra-owned or properly licensed images using approved tools
Internal concepting and exploration not intended for publication or client delivery
Generating metadata — headlines, alt text, captions, keyword tags — for assets you have rights to use
Requires Review & Approval
Any AI-generated or AI-manipulated image used in public-facing work or shared with a client, unless it fits Design's approved guidance
Any use of third-party articles, images, or creative assets beyond confirmed license terms or approved project use
Any use that could confuse ownership, originality, or attribution for client or public-facing work
Never Allowed
Inputting third-party copyrighted materials (including artists' names, artworks, trademarks, logos, or quotes) into AI tools without confirmed rights
Generating outputs that imitate identifiable branded styles, logos, or proprietary assets without authorization
Images and client-facing visuals
1
Determine whether the image is AI-generated, an AI-assisted edit of a licensed image, or a client-provided asset
2
For internal concepting, use approved tools and treat AI-generated visuals as draft concepts, not final assets
3
For client-facing or public-facing use, follow Design's approved guidance
4
Seek Legal or risk review when the use involves likeness, endorsement, trademarks, logos, third-party artwork, licensed assets, or unclear ownership
5
Keep a conservative posture for high-visibility work — maintain a record of source and license for any external deliverable
Approved Tools & Shadow AI
Use the right tools — and know how to request new ones
+
Allowed
Routine drafting, summarizing, research, analysis, and editing using approved enterprise tools
Creating Custom GPTs, Projects, or client/account assistants inside approved enterprise tools
Uploading approved internal or client materials to the appropriate client workspace, Project, or GPT
Testing new AI workflows using approved tools and approved data
Submitting new tool requests and evaluations through the AI intake process
Requires Review & Approval
Any use of non-approved tools or personal accounts for work purposes or for client work product, even if inputs seem non-sensitive
Installing or enabling browser plugins, extensions, desktop AI apps, or local model tooling on managed devices without IT approval
Never Allowed
Using unapproved AI tools with client confidential or internal confidential data
How to request a new tool
1
Submit a short request through the AI intake form
2
Specify the use case, data type involved, and risk level
3
IT reviews security and privacy posture
4
Policy owner and AI council approve, deny, or pilot with guardrails
5
AI tool terms and contract reviewed by Legal
6
Approved tools are added to the tool list and training materials
AI Notetakers
Consent comes first — always announce, always ask
+
Allowed
Internal meetings with clear consent and non-sensitive topics
External meetings where consent is explicit and content is appropriate for recording
Sharing a cleaned, reviewed summary internally — keeping raw transcripts limited to the smallest necessary audience
Requires Review & Approval
Meetings covering sensitive client strategy, personnel issues, legal topics, or crisis work
Any situation where consent is unclear or disputed
Never Allowed
Recording a meeting after a participant has objected or expressed discomfort
Forwarding raw transcript notes externally without review
On a client call with a notetaker
1
Confirm the attendee list, including any external participants
2
Use Orchestra's Legal-approved notetaker disclosure or consent language at the start of the call
3
If anyone objects, disable the notetaker for that session
4
After the call, review notes, remove sensitive items, and summarize
5
Store in the appropriate client workspace
Contractors & Third-Party Partners
Same rules apply — regardless of employment status
+
Allowed
Contractors using Orchestra-approved tools and accounts for permitted tasks — under the same rules as employees
Requires Review & Approval
Any contractor request to use personal tools, personal accounts, or non-approved services with Orchestra or client information
Any contractor request for access to Orchestra-approved AI tools
Any vendor that (a) will be inputting Orchestra or client confidential information into an AI tool, (b) uses a workflow that ingests, stores, or processes Orchestra or client materials using the vendor's own AI systems, or (c) will be using AI to create public-facing deliverables for clients — work with Legal to ensure the contract includes appropriate AI terms, and with IT and design/copy owners to ensure the right guidelines are in place
Never Allowed
Contractors using personal AI accounts or unapproved tools with Orchestra or client information
What contracts should include
1
Confidentiality requirements aligned with Orchestra's data rules
2
Tool restrictions — which AI tools are and aren't permitted
3
Disclosure expectations when AI is used in deliverables
4
Indemnification provisions for IP and data incidents
Custom AI Tools & Registry
Personal GPTs are fine — shared ones need a registered owner
+
Allowed
Personal prompts and custom GPTs used only by you — no registration required, but data rules still apply
Low-risk shared assistants (no external integrations, no cross-client data), once registered with a named owner
Requires Review & Approval
Any shared assistant intended for multiple team members — must be registered with a named owner
Any assistant that ingests client materials at scale, connects to external apps, or is shared company-wide
All integrations with outside apps — admin approval is always required regardless of risk level
Registering a low-risk shared assistant
1
Create inside approved enterprise tooling
2
Confirm no external integrations and no cross-client datasets
3
Register with minimum fields and a named owner
4
Share with the intended group and include clear usage boundaries
Submitting a higher-risk assistant for review
1
Submit through the AI intake form with the assistant description and data plan
2
IT and security review data flows and permissions
3
Policy owner and AI council approve a pilot or deny
4
Document approved scope and required review steps
5
Reassess at quarterly review

This policy is meant to help Orchestra use AI with more confidence, consistency, and speed — while protecting client trust and the standards that define the work. As tools, expectations, and use cases continue to change, this framework should give teams a clear foundation for moving forward, making smart decisions, and expanding what's possible inside approved guardrails. It will continue to evolve based on real-world usage, recurring questions, and the needs of the business.