The world is heading toward artificial general intelligence (AGI), where systems will operate in a way that’s indistinguishable from your best employees — and we’ll reach this reality sooner than you might think. 

Of course, there’s a lot of work to be done before we get there. Teams are laying the foundation for scalable AI success, but leaders aren’t always sure if their AI initiatives are working:

• 92% of companies will increase their artificial investments over the next three years, yet a mere 1% of leaders believe AI is driving substantial business outcomes today.
• Generative AI could create more than $1 trillion in value through employee productivity gains for B2B sales and marketing teams specifically, but nearly 60% of teams have yet to implement gen AI in their workflows.
• Only 31% of marketers are fully satisfied with their ability to unify customer data sources, which is essential for effective AI implementations. 

Marketing and sales leaders have a powerful opportunity — and undeniable competitive necessity — to overcome their AI hurdles and pave the path for organization-wide adoption. The key is to choose starting points that deliver measurable results with relatively little risk as they strengthen their AI strategy.  

Let’s examine what it takes to create a future-proof AI strategy for marketing and prove your ROI, including what organizations commonly get wrong when deploying AI. 

Calculating the return on investment (ROI) of AI

The principles of ROI never change regardless of what you’re measuring. The numerator represents the value your efforts drove, either revenue growth or total profit generated. The denominator is the total cost of your implementation including operational costs. 

AI investments need to deliver clear business value. And some AI use cases are more immediately valuable than others.  

When considering how to measure the value of AI projects, compare the ROI of your systems with and without it. Does the AI:

• Make you more efficient?
• Allow you to redirect your team to other priorities?
• Drive incremental revenue or profit to what you are doing today?

These considerations span many areas, some of which are clear and much easier to define. Others, not so much. 

Tangible or hard benefits of AI

Your existing processes and platforms likely measure effectiveness and results, which provides a baseline for measuring ROI of AI. For example, you may measure how efficient your people are through the number of projects or milestones completed in a given timeframe; perhaps you measure these as tickets or story points in a system like Jira. Those measures should improve following your AI implementation. 

The baseline information gives you a before-and-after snapshot. When considering the ROI of AI solutions, the metrics you want to asses may include:

• Revenue uplift - What revenue increase can you attribute to AI? Has AI increased sales through automated or improved campaigns like churn winback or cross-selling and upselling? Test individual campaigns to see where AI drives the greatest benefit. 
• Software licensing - Does your AI strategy enable you to reduce software license costs? Effective AI implementations can help reduce license costs because you shouldn’t host data in every tool or rely on channel-specific AI. AI applied at the data cloud extends intelligence across your entire ecosystem (we’ll explore this more later), so you can consolidate tools or use lower-price licenses.
• Time saved - How long does it take your team to complete tasks before and after AI? Every employee’s time comes at a cost. Get a baseline for how long tasks take before using AI by reviewing time logs or asking team members to time themselves when completing the task. Do the same after implementing AI. Use an average employee wage to generate an estimated cost per task with and without AI. Time savings can be especially pronounced with agentic AI, which operates independently and achieves specific goals autonomously while your team focuses on other areas. 

Intangible or soft benefits of AI

Other benefits are harder to measure and may be overlooked, but their importance cannot be overstated when exploring the full value of AI. 

• Change management - Does AI empower your team to optimize campaigns faster using the latest customer intelligence? Campaign testing and optimization are tricky and take considerable time when left entirely to human efforts. AI can accelerate team breakthroughs and ease transitions when shifting your strategy. 
• Cost of risk and privacy concerns - Does AI make it easier to protect customer data and enforce data governance? Improper data management practices or faulty security measures can cause data breaches or customer data misuse, which significantly hurts your customer trust. It takes considerable time and oversight for humans to enforce data governance (and mistakes are inevitable) without AI’s help. AI can be a powerful data partner, and solutions often include comprehensive data security measures and proactive alerts that help teams quickly catch and fix any issues before they spiral.
• Customer experience - How does AI empower your team to deliver a better customer experience? Is your AI helping you identify review and feedback sentiment at scale that you might have otherwise missed? This may be observed through customer reviews, Customer Satisfaction Scores (CSAT), or referrals. 
• Data preparation - How quickly does AI clean your data, remove duplicate entries, and deliver accurate campaign lists? AI is especially powerful at analyzing and organizing mass volumes of data more quickly than humans can.
• Team training - How difficult is employee training on your systems and processes? Does the AI make the process intuitive and remove frustration? Does it help people hit their stride and use the system’s full capabilities faster or access the information they need without searching for days? 

Realistic starting points for using AI in marketing

Marketing leaders should be strategic about how they start using AI to achieve the benefits we just explored. 

Leaders often think AI generates a high impact on content creation and creative tasks, but that is a very risky starting point.

Your organization has built trust with customers over time by providing a great customer experience and goods and services they believe in. While AI can augment your creative processes and make you more efficient, it’s hard to measure creative effectiveness, and any missteps could be detrimental to your customer loyalty.

Find a better starting point for AI by reviewing a dashboard your team frequently references. Choose a metric and KPI your team cares about. How can AI influence that KPI? Easily accessible opportunities may be:

• Campaign optimization - Ask AI to analyze trends in your customer data and suggest how to drive a specific campaign goal. AI can deliver recommended improvements that your team can A/B test and quickly find the right combination of elements. This is a safe starting point for AI because it can analyze your customer data and past channel messages, such as email campaigns or advertisements, to suggest improvements instead of creating new content from scratch. 
• Audience creation and lead scoring - Customer personas are the backbone of marketing success. Use AI to review your customer personas or audience lists, suggest refinements, and deliver customer lists you can target based on its recommendations. AI can also help with lead scoring based on behavior patterns to ensure the sales team targets the most viable prospects, which should drive a measurable increase in sales.
• Competitive messaging analysis - Competition is fierce in every industry. It’s easy for teams to fall behind on the new entrants in their space. Use AI to analyze your company messaging based on your audience's needs. Ask it to also identify competitor offerings and describe those brand platforms to find areas where you can stand out and improve your competitive standing.

Scaling your marketing success with AI

Introducing AI in easily measurable and safe areas will help you prove your initial results. However, there are many roadblocks you may face when expanding your scope or achieving the full potential of AI.

Use cloud-native martech for fully inclusive AI

For AI to fulfill its potential for your organization, it needs to know about your customers and your history with them. 

You’ll want to have control over exactly which data the systems can access in a fine-grained way based on opt-ins, opt-outs, and other privacy considerations. Over time, you’ll likely give the systems access to more customer data as you build safeguards within your system and trust within your organization. For both control and breadth of access, there’s one place that makes the most sense to integrate your AI: the data cloud. 

You should apply AI to your data cloud because it is the highest upstream system, with a single source of truth for your customer data. Whether you’re deploying ads, emailing customers, or actually calling customers, the data cloud stores all information and is the ideal location to build audiences and optimize their journeys. 

Composable martech is built around your data cloud as the single source of customer data for all channel tools. This cloud-native approach is a good starting point for measuring the impact of AI because all data and results sit in one location. This allows you to measure holistic results instead of channel-specific and siloed insights, empowering you to assess vendors and optimize your martech stack for cost savingswhile boosting performance. 

Avoid vendor lock-in

Don’t bet your entire AI strategy on one vendor. Look for solutions that connect to your data cloud and allow you to use your own fine-tuned or hosted AI models, in addition to the vendor’s large language model(s), if you want. 

By connecting to the data cloud, you can quickly swap best-of-breed frontier models or vendor solutions that leverage them — without having all of your business logic locked up in a single marketing cloud, CDP, or other suite that restricts access to your own business data. If you make a bet on one vendor for a long-term contract and suddenly there is an exponentially better application of AI available, you could find yourself in a tough spot. 

When assessing vendors, ask questions like:

• How accessible is the solution? Can an engineer access the system to do further analysis?
• Are the artifacts your system creates stored in my data cloud? Can I export them easily if not?
• Can we use our own models or fine-tune the provided models? 
• How does this solution connect to my other martech? Can I use the assets or business intelligence we generate in your system across every surface I use to communicate with my customers?

Balance oversight and flexibility with AI committees

A lot of companies we work with have AI committees, and they offer different levels of value. Too often, committees inadvertently stifle innovation by inviting too many members without a clear purpose, or they evaluate solutions for longer than necessary because they lack a clear idea of what AI should accomplish for the organization.

Committees are most valuable when they set up AI principles, prioritize implementations, and ensure everyone’s input is reflected in the strategy. 

Don’t sweat the costs of inference 

You shouldn’t worry too much about the cost of using AI models when calculating your AI ROI. 

Of course, be responsible and set quotas for tokens or spend per day or month, but don’t consider today’s inference costs as a constraint. The cost per token for the same level of intelligence is decreasing by 10x per year. 

While scaling laws on the pre-training side are starting to hit diminishing returns, scaling of inference time compute (i.e., “reasoning”) should continue to exponentially increase intelligence for some time. 

Be sure to skate to where the puck is going and not limit your design space by believing the costs per token of today will be as high as they are six months from now. 

This applies to your vendors as well. If they are passing through your usage costs, ask if the costs per token are variable or fixed once you sign the contract; this could be the difference between an ROI-positive project and one that costs more than it needs to.

Unified-loop marketing drives ongoing AI improvement

Google Cloud CMO Alison Wagonfeld often discusses going from the “wow” to the “now and how” of AI. While it’s easy to get distracted by enticing demos — which are conducted in very controlled, boxed-in environments — organizational leaders need practical solutions to implement AI in ways that guarantee their ongoing success.

• Your data cloud and organizational intelligence are your best differentiators. Ensure all data is stored in your cloud and apply AI there to drive efficiencies across any channel tool. 
• Start by focusing on a single metric you can influence with relatively little risk. Use controlled use cases to prove the initial results of your AI investment. 
• You can then find new areas to expand and augment what your team can accomplish.

Setting this foundation enables a unified loop for innovation, where real-time insights deliver seamless personalization across every customer touchpoint. By using AI in the right places, your team can accomplish more complex tasks and make informed decisions that move your business forward.

Don’t let uncertainty stop you from getting started.