Getting Started with Enterprise AI: Part 1

First, a disclosure. 

Of course I used AI to assist me in the composition of this three-part blog series. Now and increasingly moving forward, fewer and fewer content pieces will be free from some use of AI on the work. These are undeniable productivity tools. It is my firm opinion that those who choose not to leverage AI in some capacity remain like the scribe who scratches out equations on paper when the calculator is sitting right beside them. 

That being said, this is not an article on auto-pilot. And I am confident that the good majority of you, having interacted with AI content tremendously at this point, can tell the level at which a piece is automated as opposed to developed in collaboration with human effort. The structure of content I present is sourced completely from my human strategic creative center. While AI has assisted me in paragraph expansion and narrative flow, the approach which I lay out is born from my deep experience of working with clients and the vision I can more effectively implement through the tools I now use to skyrocket my productivity.

Introduction

Most articles about artificial intelligence provide generic overviews of capabilities and use cases without practical strategy for using the tools in the context and environments in which we work. 

But to truly embrace the benefits and value of this radical technological pivot point (in which the hype is fully justified, in my opinion), enterprise creative & marketing teams need specific, tactical advice on implementing and adopting AI operations (AI Ops) which balances ease of entry with organizational privacy and other concerns.

That’s why this article is different. It delivers an actionable blueprint to lay the foundation on the use of AI within your organization to start benefiting from its value appropriately and can transform readers from "AI-curious" to "AI-empowered".

My goal in this article is to provide you with:

• A step-by-step framework for building the foundations needed to effectively scale AI adoption across teams

• Starter use cases that offer quick wins like basic content creation and design generation  

• Awareness of the importance of an AI Council for governance

• Actionable steps for launching an initial pilot

I believe the pragmatic approach I’ve outlined will enable readers to lay the foundation of an AI Operations practice in their organization so that they may begin to experiment with intention and structure to realize the gains promised in this revolutionary technology.

What is AI Ops?

AI operations (AI Ops) refers to the application of artificial intelligence technologies to enhance and optimize business operations and workflows. It focuses on using AI to assist humans and augment human capabilities, rather than replace them.

For creative and marketing teams, this means leveraging AI tools to help with:

• Automating repetitive administrative tasks

• Producing content variations like text, images and video

• Optimizing assets for different platforms  

• Analyzing campaign performance data

• Providing insights to inform strategy

The goal is to use AI to empower human marketers and creatives, not replace them. AI handles executional work while people provide oversight and judgment.

The Foundational Components

Before jumping into experimenting with AI, it's incredibly beneficial to invest time up front to establish the following foundational components. Without these in place, you're herding cats. Individuals may advance at their own paces, but the organization will not realize the benefit of a cohesive program as a whole. These areas of focus provide the structure and governance needed to support your AI initiatives.

Project Management

To govern and gain visibility into AI operations work, it’s important to track activities in your project/work management system. Rather than introduce an entirely new platform, I recommend creating a separate portfolio to house all AI-related initiatives and pilots so that these activities can be segmented, but still reported globally against other BAU initiatives. Tracking AI projects distinctly allows you to isolate progress and performance for clear reporting and insights. 

It’s important to document key details on each project such as goals & objectives, the source of data input, AI tools used, target metrics, risks, and results. You can also apply known prioritization frameworks, such as WISE (see below), to assess and prioritize opportunities based on workload, investment, savings potential, and level of effort required.

W - Workload - How frequent is the task?

I - Investment - What are the costs to perform it?

S - Savings - What efficiency gains from automating?

E - Effort - How complex to automate with AI?

(The aggregated score of responses across these questions can provide clear awareness of which projects to consider automating.)

Metrics Strategy

To quantify the impact of AI, it’s important to clearly define key metrics aligned to goals like time savings, output volume increases, and asset creation time per campaign. Tracking operational performance data informs strategic decisions and helps make the case for future AI investments

For example, establish benchmarks for:

• Time spent on routine tasks before AI automation

• Volume of content types produced weekly/monthly

• Number of iterative cycles or versions per campaign

Categories of Metrics to consider include:

• Prompt-level metrics - to inform the efficacy of the prompt engineering strategy

  • Examples: Prompt relevance score, prompt usage frequency, prompt iterations

• Operational metrics - to inform efficiency gains and productivity

  • Examples: Time savings on tasks, volume of content output, asset creation time

• Content-level metrics - to inform the performance of actual creative assets

  • Examples: Engagement, conversion rates, audience sentiment

• Pilot metrics - to inform the success of AI implementation

  • Examples: User adoption, tool ROI, stakeholder feedback

Prompt Library

First, what is a Prompt?

A prompt is the text input provided to an AI system that instructs it on what kind of output to generate. Prompts include details like style, aspect ratio, camera type and purpose which are used to steer the AI response. Crafting effective prompts is key to producing useful AI content.

What is Prompt Engineering?

Prompt engineering refers to the iterative process of carefully crafting, testing, and refining prompts to improve the relevance, accuracy, and consistency of AI output. It is both an art and a science.

The value of a Prompt Library

Any time you use generative AI, you're effectively rolling the dice on the output. A shared prompt library lets teams capture proven prompts that reliably produce on-brand content, saving others from reinventing the wheel. Continually optimizing prompts through new examples and performance data maximizes consistency over time.

Establishing a robust prompt library is essential for managing your AI tools and components. This centralized database allows you to catalog and organize the prompts used to guide AI content generation in each tool, as well as track results to improve outputs over time.

Start by developing a taxonomy to categorize prompts by use case, content type, or other attributes. Document prompts using a standard template that includes fields for prompt text, parameters, AI tool used, and measured performance metrics. Tag prompts based on factors like tone of voice or goal outcome for easy filtering and reuse.

Over time, analyze performance data to refine prompts and usage guidelines. Remove biased language. Add new proven prompts and share learnings across teams. Think of your prompt library as a living asset that grows in value through iteration and collaboration.

Example Taxonomies for Organizing Prompts:

• Campaign Stages: Brief, Concepting, Asset Development, Approval, Launch, Reporting

• Asset Types: Display Ads, Social Posts, Emails, Landing Pages, Videos

• Campaign Elements: Titles, Taglines, Copy, Images, Layouts

Example Metadata Fields to Track Prompts:

• Brand Fit Score: Quantify how on-brand the asset is

• Tone Rating: Aligns to brand voice (fun, serious, emotive, etc.)

• Channel Optimization: Score asset optimization for platform

Digital Asset Management

To effectively manage AI-generated content, consider expanding your approach to Digital Asset Management (DAM) with metadata and organizational decisions to specifically track the content generated from AI initiatives (or with the use of AI).

For example, you may want to:

• Update metadata fields to capture key data on AI usage including prompt ID, AI model used, and % generated by AI vs. humans

• Build collections, tags, and taxonomies to identify and isolate AI-influenced assets

• Integrate your prompt library metadata with your DAM to link prompts to resulting assets

With the right metadata foundation, your DAM can become the command center providing visibility into AI's impact on content production across campaigns and teams.

AI Operations was one of several hot topics presented at Cella’s signature professional development conference, Illuminate. This annual event presents transformational thought leadership and one-of-a-kind instruction from our team of experts, notable names and leading experts in the industry. Read more on our Illuminate discussions here in our retrospective blog. Learn more about Illuminate by Cella here.