Artificial Intelligence

Generative AI: How Close Are We to a Crossover Point?

When Toys”R”Us debuted their Sora AI generated ad last week, it set a new benchmark in terms of how quickly we are approaching a Generative AI media future. While the ad itself shows that the underlying technologies still have quite a bit of improvement before they’re ready to generate high fidelity media, it is nonetheless an achievement and a building block for future use.

What does this mean for the future of creative work and how will brands decide what to shoot in real life and what to generate instead?

What is the AI Crossover?

The current creative workflow is to draw, shoot, record and model out creative before using computers to then edit, correct and sequence assets into a final format. The crossover is the point where the majority of media (static images, videos) is generated by AI first and then subsequently edited by humans into a final format.

That crossover represents both a technology and confidence hurdle for generative AI and in the Toys”R”Us ad it’s clear that a lot of post-production work had to go into the final product just to make it function. Beyond generative issues with perspective, the clip featured noticeable edits that likely weren’t part of the original generated creative.

For a brand that has access to strong post-production teams, the generative AI crossover will likely be sooner rather than later as they can afford to work with middle of the road assets to build a completed product. However for smaller brands, the level of fidelity needs to be much higher before they can confidently begin building in AI.

Thus while adoption will continue to accelerate in some areas, the technology is still not at a level where it can universally meet the creative needs of all brands.

Cost to Shoot vs Cost to Generate

Once the technology gets to a point where generating high fidelity creative is possible, then the question as to whether creative will follow a traditional workflow vs a generated one will come down to cost.

Depending on the provider, it can cost anywhere from $0.05 to $0.12 to generate a single image based on token pricing. If you are trying to create a 30 second commercial at 30 frames per second then your rendered cost is less than $100 but that doesn’t include the number of iterations that would need to be made before you have a final product in addition to any renders you make for storyboards, concepts etc …

I would venture to suggest that the cost just to render a 30 second ad would easily be $4-5k all in and that’s before you add on any additional costs for pre/post production, sound, music licensing, or human voiceovers. Thus the true cost would be a multiple your render costs based on how much additional work is needed.

Given these parameters, brands looking to invest in Generative AI for use throughout their creative ecosystem should begin documenting and comparing costs between traditional ad creation and generative to determine which will be the better use in different scenarios.

Starting with the medium, audience and use brands should develop T table analyses of traditional costs like talent, locations, crews, pre/post vs pure rendering and editing to begin creating decision trees for their projects.

AI is Workflow Ready

While Generative AI may not yet be the right tool for producing final assets, it’s clear that the technology is ready for daily use as a way of creating sketches, proof of concept and storyboard pieces that can be used to inform high fidelity creation.

I know there is a lot of concern about the impact of AI on creative fields but at the very least adopting it as a workflow solution as an individual, agency or department will produce better returns in the long run.

Furthermore creative minds are still an asset that no amount of prompting and prodding will ever be able to replace so even if we get to a point where a 30 second commercial can be rendered in nothing more than a set of prompts, a mind and vision will still be required to create it.

Originally posted on my LinkedIn

Artificial Intelligence

The Business Case for Chief AI Officers

Creating a centralized role and subsequent office to manage AI technology initiatives and investments is the single best strategic move that any enterprise will make over the next decade.

A single authoritative office that can help departments ideate use cases, define requirements, review contracts, own the global AI roadmap and provide organizational thought leadership will be critical to ensuring AI technology is adopted as efficiently and effectively as possible.

Let’s examine how this cross-functional team would own AI throughout an organization and why it should be its own department within a corporate structure.

Governance and Roadmap Ownership

Key tasks of the AI department would be owning and developing the global roadmap in addition to setting governance for AI projects.

Having an enterprise roadmap is an absolute must because without it, organizations will inevitably push and pull efforts in numerous different directions without an understanding of effort and overlap. Tasking the AI office to create, revise and cast the global roadmap is critical to aligning all pieces of an organization, forecasting costs and providing clear timelines for innovation.

Second, and arguably just as important, is having the office own AI governance for the organization. Governance doesn’t get anywhere near as much attention but every organization needs a strong governance facility to set rules, ensure compliance and make sure professional and ethical considerations are taken into account when building, licensing and launching technology.

Frontline Organizational Support

The second major area of responsibility is for this office to be a day-to-day clearing house for all AI projects. Consider these three use cases:

1) R&D wants to build a proprietary LLM for research but doesn’t understand the significant development costs, data utilization and data cleanup required

2) HR wants to upgrade their internal knowledge base capabilities with a RAG system but doesn’t know whether RAG systems and data calls pose any compliance risks.

3) Marketing wants to build their own Generative AI tool but is unclear how much material is needed to train models on their products and whether they will have copyright on the output

In each scenario, a lack of guidance from an authoritative source can create liabilities, inflated budgets, poor product performance and potential regulatory violations.

Take for example R&D looking to build a LLM – assuming they have a vetted use case, creating your own LLM requires significant data storage, data cleaning, server utilization costs and other expenses of which a team may only have a limited understanding. A $2 million budgeted project is more likely $20+ million before you even have a working product.

In this case, the R&D team being able to share the use case and corresponding plan with an authoritative internal team would immediately correct estimates, align the project within the roadmap and if necessary provide a collaborative business case for the full investment.

These exact situations are already happening throughout global industries and a lack of concise guidance is the difference between a successful product and wasting millions on a partially baked idea.

Why Not Have AI Live Under CIO/CTO?

AI has strong IT and technology components and one could make an argument for putting it under either department but I feel independence is critical given the power that AI will wield in the next 10 years.

If AI exists solely under CIO, it will be viewed through the lens of IT as 80% implementation and 20% vision. Whereas if it exists under a CTO, it will be 80% vision and 20% implementation and neither is optimal when you need both considerations in balance.

Artificial intelligence will transform how operations are conducted throughout all departments of an organization but not without significant upfront investment. Having a neutral clearing house that can provide objective assessments, in a race where speed to market is the main factor, is critical to avoiding costly mistakes that can set you back years.

Who is the Ideal CAIO Candidate and How is it Structured?

Ideally you need a generalist who can pull from a deep pool of experience and understands the underlying technologies and data requirements for AI to function successfully. They should then have a staff to lean on with a strong supporting team that can provide global perspective on regulatory, contractual, ethical and practical considerations.

The team should then connect to VPs from the CIO and CTO layer to form an internal council that is entrusted with owning, approving and guiding AI within the organization.

In the Interim

While enterprise organizations are still setting up nascent AI practices, having outside counsel that can objectively look at the whole of your company’s artificial intelligence efforts and provide support for ongoing and planned projects is absolutely critical.

Organizations are already investing heavily in AI but often at the whim of who is able to advocate the loudest for their project as opposed to following a global plan. Having a partner like TCS who can provide leadership and authority to enterprise AI can align efforts and ensure effective investment.

(I originally posted this to LinkedIn and reposted here)

Photo credit: Nathan Anderson / Unsplash