Images, infographics, posters, banners, forms — and now Sora 2 video. The Copilot app's Create tab has quietly become the most visually capable AI tool in the enterprise. For advisory teams used to commissioning visuals, this is a category shift. Used carefully, a major accelerator. Used carelessly, a governance exposure. This chapter is about the careful use.
Before the features, the reframe. Visual content in advisory work used to live at the end of a commissioning chain — design brief, designer, revision, final asset. The Create module collapses that chain for a specific class of visuals. Knowing which visuals belong in the new pattern and which still need the old pattern is the core skill.
Think about how visual assets actually get made at Arup today. A consultant needs an infographic for a client briefing. They either: (a) wait for the design team's queue, (b) cobble together PowerPoint shapes in a way that looks acceptable-but-homemade, or (c) commission something external. The cycle is days or weeks for option (a), the quality is mediocre for option (b), and the cost is measurable for option (c).
"The quality gap between 'AI-generated' and 'designer-produced' continues to narrow, and the speed improvements make Copilot even more viable for rapid ideation."
The Create module, paired with the latest image model (GPT-Image-1.5) and OpenAI's Sora 2 video model, collapses specific categories of this work into a 60-to-180-second loop. Three shifts matter:
If the asset you need is a standard type — an infographic with known content, a branded poster for an internal event, a banner for a newsletter, a poll-form for a client survey — the Create module produces it in one to three minutes. Faster than most Arup consultants could produce it manually; often comparable in quality to commissioned work when the brand kit is properly configured.
The older failure mode: consultants gave up on including visuals because they didn't have time or capability. The new pattern: generate a first attempt, iterate with prompts, arrive at something usable. Lowering the floor on "can I make this myself?" from "no" to "yes, with three iterations" changes what consultants attempt.
Not every visual should now be AI-generated. The new question is: given the audience, purpose, and stakes of this asset, which path does it belong on? Internal team comms poster? Create module is fine. Board-facing strategy deck diagram? Probably Create module with brand kit, reviewed by lead. High-stakes client-facing brochure? Still the design team. The skill is knowing where the line sits — and the lines are different from where consultants instinctively draw them.
For Arup specifically — where the brand is a competitive asset and the advisory team's deliverables carry explicit quality expectations — the Create module is best used as an accelerant, not a replacement. The partners who get the most value from it are those who treat it as a new tool that fits certain categories of work, not as a blanket solution.
The Create tab surfaces eight content types but six strategic capabilities for advisory work. Know what each is for; pick deliberately.
config.office.com), adds a security group of designated brand managers, and those managers create and publish official kits. Official kits become available to all tenant users; personal kits can be created by individuals but aren't shared. For Arup specifically, the ideal state is one or more official Arup brand kits configured by the brand team — primary Arup identity, sub-brand variants (if any), sector-specific templates. Without this setup, the Create module produces output that looks competently AI-generated but does not look Arup. This is the single highest-leverage Create-module admin task.A dedicated section because it matters. AI-generated visuals aren't neutral — they carry implications for brand integrity, client relationships, disclosure, and intellectual property. Knowing where the green, amber, and red zones sit saves consultants from learning through uncomfortable client conversations.
AI-generated content used within Arup, for Arup audiences, where AI origin is implicit or explicit.
Assets that may reach external audiences but where scrutiny is lower, or where the asset is ancillary to a human-authored deliverable.
Assets where AI origin could damage trust, where brand integrity matters commercially, or where inaccuracy carries real risk.
A simple test for any borderline case: would you be comfortable if the client, reader, or audience knew this was AI-generated? If yes, proceed. If the answer is "probably no, but they probably won't ask," that's a red flag — use an alternative. If the answer is "we'd need to tell them," add the disclosure and proceed. The discomfort with disclosure is the signal.
Sora 2 videos come with a visible "AI-generated" watermark that is unavoidable by design. Don't try to crop, cover, or remove it. The watermark is a feature — it means Sora 2 videos are self-disclosing, which reduces governance risk and matches the broader direction of enterprise AI governance (regulators and professional bodies increasingly expect AI-generated content to be identifiable).
Unlike Sora 2 videos, AI-generated images do not carry a visible watermark. This places the disclosure burden on you. Practical guidance:
AI-generated imagery in the UK (and in most jurisdictions Arup operates in) has unclear copyright status. You can use it; you may not reliably own it in the traditional sense; competitors can use similar models to generate similar outputs. For Arup-proprietary visual assets that need defensible ownership — brand marks, published materials, signature imagery — use human-created content. For ephemeral visuals (internal communications, workshop materials, LinkedIn posts) the IP question is academic.
This one matters more than it might seem. Arup's brand distinctiveness is part of its commercial value. If the team generates a lot of AI imagery that looks generic-Microsoft-AI — because nobody applied the brand kit, or there was no brand kit configured — the cumulative effect is a dilution of brand distinctiveness. Consultants individually don't see the harm. Cumulatively across hundreds of pieces of content, it's real. The discipline: always apply the brand kit, or don't use AI-generated imagery for the asset. Brand kit off = generic; brand kit on = Arup-identifiable.
The Create module works out of the box for anyone with an eligible M365 subscription — but output quality and appropriate use depend on the brand kit being set up and the team understanding the governance reality.
1. Confirm your licence tier gives you the full Create module. M365 Copilot (Premium) unlocks the full experience — all content types, Sora 2 (when Frontier), brand kit integration. Copilot Chat (Basic) users may see a reduced Create module. Check by opening the Copilot app → Create tab and confirming you see images, infographics, posters, banners, video, and forms as options.
2. Check whether Arup has an official brand kit configured. In the Create module, when you generate a poster/banner/infographic, look for Brand and color. If an Arup brand kit appears, it's been set up. If only generic colour palettes appear, it hasn't. If not set up, this is the single most impactful ask to make of IT and the brand team — without it, AI outputs look generic-Microsoft-AI rather than Arup.
3. Check Sora 2 availability. Sora 2 is Frontier-program-only. Check if Arup is enrolled in the Frontier program by going to Create → Create a video → describe a video and see whether Sora 2 appears as an option. If Sora 2 doesn't appear, only stock-footage-based video generation is available. Frontier enrolment is a tenant-level admin decision.
4. Know where outputs are stored. Created content saves to your Copilot Library (for images, posters, banners, infographics) and to OneDrive for Business (for videos, including Sora 2 output). All outputs are subject to full Microsoft Purview governance — DLP policies, sensitivity labels, eDiscovery, audit logs — same as any other M365 content. This matters: AI-generated content doesn't escape the governance perimeter.
5. Brief the team on the green/amber/red zones. The single most important training step is not "how to use the Create module" but "when AI imagery is and isn't appropriate." A 30-minute internal briefing covering the three zones, the disclosure test, and the IP question prevents most downstream mistakes. Without that briefing, enthusiastic early adopters will produce a client proposal with AI imagery that causes an uncomfortable conversation, and the team learns the hard way.
The highest-value default use case for advisory teams. A client briefing goes out each week. Attached to it, or bundled into the slide deck, an infographic summary lifts the comprehension and the perceived polish. Historically too expensive to commission for every brief; now a two-minute task.
Generic prompts ("make an infographic of this") produce generic output. The difference between a usable infographic and one that needs to be discarded is in three specific prompt moves:
Every AI-generated infographic needs a fact-check edit pass before it leaves your desk. Copilot may have:
Click Edit after generation; the editor opens. Walk through every element against the source document. Fix text errors directly; adjust icons if they're confusing; reorder if needed. This edit pass typically takes 3-5 minutes and is the difference between a usable deliverable and a liability.
1. The client briefing executive summary. One-page infographic of a 3-10-page written brief. Accompanies the brief as a visual summary. Portrait format. Audience is senior client stakeholders.
2. The framework diagram. Visual explanation of a methodology, process, or framework. Used in proposals and internal explainers. Wide format often works better. Audience varies.
3. The timeline / roadmap. Programme phases with milestones. Wide format. Accompanies strategic proposals and programme kickoff briefings.
4. The comparative one-pager. Options analysis, competitor comparison, framework comparison. Typically wide format. Useful for slide decks and client options papers.
The green-zone use case. Internal-only audience; AI origin is fine; brand kit carries the heavy lifting. Team event announcements, channel banners, campaign launches, training invites — the assets that used to take design-team time or just didn't get produced.
For high-volume internal-comms use, build a one-minute muscle-memory pattern:
Even in the green zone, three quality checks before deploying an internal asset:
Five-second check per asset. Catches the 10% of outputs that would otherwise embarrass the team.
The use case that lives outside the Create module but is enabled by the same image-generation stack. PowerPoint's Copilot now generates context-aware imagery inline — look at the slide's content, generate imagery that actually fits, refine without leaving PowerPoint. The quiet PowerPoint enhancement that changes how slide decks get made.
Part V covered PowerPoint Copilot in depth. The context-aware image generation sits inside that, but it's worth highlighting specifically here because it's different from the Create module's image generation:
Rules of thumb:
Brand kits (Section 2, tool card 6) work for PowerPoint image generation the same way they work for the Create module. With Arup's official brand kit configured, PowerPoint generates imagery that respects Arup's visual identity — colour palette, imagery style, compositional conventions. Same argument as in Section 4: the single highest-leverage setup action is getting the brand kit configured. Once done, it benefits both the Create module and PowerPoint image generation.
In the past year the quality gap between AI-generated presentation imagery and human-commissioned imagery has closed meaningfully. GPT-Image-1.5 (January 2026) was a notable step up from its predecessor. Images now look plausible as designer-produced for many use cases. But designer-produced imagery still wins for: anything depicting real Arup projects or people; anything requiring specific technical accuracy; anything that will be scrutinised for brand integrity; and hero imagery for formal client materials where distinctive design thinking matters. AI-generated imagery wins for: volume of illustrative imagery; rapid iteration; internal and low-stakes use; imagery needed at speed.
The capability most different from everything else. Generate a short video — realistic motion, synced audio, cinematic or realistic styles — from a prompt. Currently Frontier-program-only. Transformative for internal comms; explicitly not for client deliverables without explicit disclosure.
The model has four capabilities worth knowing:
Every Sora 2 video has a visible "AI-generated" watermark overlaid on the clip, provided by Microsoft/OpenAI. You can't remove it. This is the feature, not a limitation — it means Sora 2 videos are self-disclosing. Any attempt to crop, cover, or obscure the watermark violates the terms of use and defeats the governance rationale.
Sora 2 video belongs firmly in the green zone (internal use) and nowhere else until Arup's brand team has explicitly blessed external use. Some defensible patterns:
1. Internal training micro-videos. A 15-second clip demonstrating a concept in an onboarding module. AI origin clear; no client-facing implications; speed of production makes training authoring viable that wouldn't be otherwise.
2. Internal campaign hook videos. The 10-second teaser for a practice-area initiative, shared in the team's Teams channel. Watermark visible; audience knows it's AI; purpose is engagement not deception.
3. Concept visualisation for brainstorms. Before commissioning real production, generate a rough Sora 2 video to help the team visualise what they're after. Concept tool; never the final product.
4. Internal milestone celebrations. Short celebratory clip for a team all-hands or project close. Watermark is part of the charm; no external audience.
The under-the-radar Create module option. Microsoft Forms has been available for years; Copilot's prompt-to-form interface removes the setup friction. For advisory teams running workshops, pre-reads, stakeholder input, and quick internal research, this collapses "I need a form" from "an hour of Microsoft Forms clicking" to "describe what you need."
Form prompts benefit from explicitly specifying four things, because Copilot makes better choices about question types and logic when it knows them:
AI-generated forms have consistent failure modes worth knowing:
The strongest advisory use case: pre-workshop or kickoff-meeting questionnaires. Traditional pattern: facilitator creates questionnaire (30-60 min), distributes, collates responses manually (1-2 hrs), references in session. New pattern: generate form (2 min), distribute, Excel collates automatically, 15 min to review responses before the session. Recovers 2-3 hours for every significant client workshop. Across a year, this is a material productivity gain.
A reminder worth stating directly: forms collecting personal data are subject to GDPR and Arup's own data handling policies. If you're collecting genuine PII (names, roles, contact details, opinions attributable to individuals), the form itself needs appropriate notices and the data needs to be handled through Arup's formal processes — not left in a consultant's personal OneDrive indefinitely. For anonymous feedback forms, lower governance threshold. For anything identifying individuals and their views, check with the engagement manager or data-protection lead before distributing.
The use case that isn't a use case — it's a structural prerequisite. Without a properly-configured official Arup brand kit, every other Create module output is generic by default. With it, every output is Arup by default. This is the single highest-leverage action in the chapter, and it belongs to the brand team and IT more than the advisory team.
Step 1 — IT admin configures the Enterprise Brand Manager policy. Via config.office.com, the Office admin creates the policy and assigns a security group containing the designated brand managers. Up to 24 hours for permissions to propagate. One-time configuration; then changes only if membership changes.
Step 2 — Brand team curates the content. The designated brand managers (marketing / brand team members) gather Arup's approved brand elements:
#E61E28, dark navy for text, white and light grey for backgrounds, any approved accent colours. Specific hex codes.Step 3 — Brand managers publish the official kit. In the Copilot app, brand managers with the Enterprise Brand Manager role see a Publish option when creating kits. Publishing makes the kit available to every user in the tenant. Multiple kits can be published — primary Arup identity, sub-brand variants, sector-specific kits (transport infrastructure, buildings, cities), if Arup wants that degree of differentiation.
Step 4 — Advisory team selects the kit in every Create module output. From the advisory team's perspective, the only change is a habit: when generating a poster, banner, infographic, or image, select the official Arup brand kit from Brand and color before clicking Create. Taking it for granted that this single click is the difference between Arup-looking and generic-looking output.
Step 5 — Periodic review and update. Brand kits need occasional refresh — new brand guidelines, additional sub-brands, sector-specific kits. Brand managers can edit, rename, or delete. Typically a quarterly review cadence is sensible; major refresh only when the brand itself evolves.
Every other use case in this chapter depends on this one working. Without a configured brand kit, a consultant generating a client briefing infographic (UC1) gets something that looks competently AI-generated but doesn't look Arup. A consultant designing an internal event poster (UC2) gets a generic corporate aesthetic. A consultant using PowerPoint's image generation (UC3) gets context-aware imagery that nonetheless doesn't respect Arup's visual identity.
The brand kit is the single input that makes the Create module actually useful for an organisation with a brand. Without it, the team is either producing off-brand content or manually fighting every output to bring it back on-brand — which destroys the speed advantage entirely.
Three parties need to coordinate:
None of these is the individual consultant's job. But without all three doing their piece, the individual consultant cannot produce on-brand AI imagery, and the rest of this chapter's value doesn't materialise.
Six use cases, six surfaces, one table.
| Your scenario | Use this | Surface | Time |
|---|---|---|---|
| Client briefing needs a visual executive summary | Infographic · upload brief, generate | Create module | ~5 min |
| Internal event announcement poster | Poster · brand kit on | Create module | ~2 min |
| Teams channel banner for an engagement | Banner · brand kit on | Create module | ~2 min |
| Slide deck needs illustrative imagery | PowerPoint context-aware image gen | In PowerPoint | ~1 min / image |
| Pre-workshop questionnaire for client team | Forms · describe purpose + questions | Create module | ~3 min |
| Short internal campaign hook video | Sora 2 video (if Frontier) | Create module | ~5 min |
| LinkedIn post imagery for thought leadership | General image generation | Create module | ~2 min |
| Framework diagram for a proposal | Infographic · wide format | Create module | ~3-5 min |
| Stock video for internal comms (no AI origin) | Create a video · stock footage path | Create module | ~3 min |
| Formal client-facing marketing material | Design team, not Create module | Offline | n/a |
| Images of Arup people, offices, or projects | Real photography, not AI | Photo library | n/a |
| Rapid concept visualisation for brainstorm | General image gen or Sora 2 | Create module | ~2 min |
Patterns from training advisory teams on the Create module. Almost every bad outcome maps to one of these.