Marketing technology is changing. In the course of my career, I have watched various technology trends emerge, grow, or fade (Ana Mourao does a great job explaining the three ages). First, we jumped into the world of marketing platforms to ensure all touchpoints were centralized in a single, marketing-owned database. Then we watched the technology offerings jump to 15,000 products. So, Marketers have been able to build their own apps with low-code/no-code products. And the script flipped with the introduction of generative AI (where possibilities have become endless).
Where the past seemed like a linear change, today we are in pure chaos. There is so much hype in the marketplace that making decisions seems impossible. In this post, I will go through what I am observing and how to address it.
The Marketing technology Vendor Landscape is Constantly Shifting
“Have you heard of the xyw tool? It’s amazing; you should check it out.” Is a common comment I am hearing from my peers who are experimenting with new tools. And there are so many new players in the market.
As I was building out the base of my stack, I started to look into additional data enrichment tools and other tools to help me prioritize my conversations. This is where I started to notice a concerning trend. I was first told to check out the GTM engineering tool for data enrichment. The buzz around this tool was crazy from December to February. Then the same folks who recommended that tool in March and April told me that the agentic agent builder tool was now the way to go.
Wow, within three months, the best-in-class marketing technology tools changed. In addition to changes in the features and functions of tools, there is consolidation and pivoting among Marketing technology vendors. When you are starting a vendor onboarding process, and by the time you complete the process, the organization has been acquired (this happened to me once). If you’re a small business, not a big deal. But if you are a mid to large-sized organization, purchasing software can take anywhere from three to twelve months to complete. The business case for purchasing the tool must be solid, the information security reviews must be completed, and implementation and training must follow. Because of all the steps leaders must go through, there is no backtracking. I can understand the low confidence that marketing leaders feel today when purchasing technology. If you’re able to take short-term contracts to test and learn from tools, now is the time.
Tokens vs Subscriptions
Marketing Technology is available to purchase on a Subscription as a service (Saas) model. The buyer will subscribe to receive access to the cloud-based software. The buyer pays a flat annual fee. The annual fee includes the cost of access to the software, the number of users, and other potential add-ons. Consumption-based pricing charges per action, query, or unit of output. Understanding this distinction is critical because it changes how you budget, forecast, and evaluate ROI.
A new trend has been spreading in the technology world: consumption-based pricing. Word on the street was that this idea arose because executives complained that they paid for large subscriptions yet were not using the full set of tools. Software is a very low-cost business to build. You need a server, write code, lease the software to users, and keep updating it in the cloud. Moving to the consumption model, now providers can charge for each function within the software they provide to you.
There are a few concerns we should acknowledge about this consumption approach, including learning the tool and managing costs. With token-based models, users must pay for each action they take in the tool. Education and learning are not excluded from token cost. To learn a token-based tool means you must pay for the actions you are learning. This is a very expensive learning curve. In the past, you could access a tool and learn without fear of financial implications.
Let’s do the math.
| Subscription A Marketing Software for 5 user | Consumption Subscription* A Marketing Software for 5 users |
Total: $15,000 annual subscription before taxes | $199 per month per user & 40,000 credits per month, overages $1.00 per credit $199 x 5 =$995.00 for 5 users $995 x 12 =$11,940.00 annually Overages = Between $40,000 – $200,000 (double credits per user) Total: $51,940 – $240,000 annual before taxes |
*Companies that offer tokens are not clear about which functionality requires tokens. Not understanding which features require additional fees makes it harder for the manager to budget accurately for the subscription cost. Assuming each user will use 40,000 credits, the token purchases must also reflect this.
This is a one-year view of the cost. Know that this tool will likely need to remain in your stack going forward. So, looking at the cost year over year to understand the impact this tool will have on your budget annually. Most SaaS companies now require a 5%-10% annual increase in the subscription. It is unclear what the token and token functionality needs will be with the new tools.
Managing the marketing technology budget going forward will be challenging. A lot of the AI products look cheaper today, but they are at introductory pricing. The underlying technology and infrastructure needed to power AI are costly. One would assume that, since operational costs are higher, service costs will increase, allowing those AI companies to become profitable.
Build vs Buy
Building with No Code
In 2021, we saw the rise of no-code tools being adopted in the marketplace. I, for one, have tried out both Salesforce and Microsoft no-code functionality. Knowing you could build an app on those platforms that met your requirements was cool. As an example, a colleague and I built a UTM parameter tool on Microsoft. We wanted something more than Excel, but we needed the history of past URLs with UTM. We wanted consistency with naming conventions in the UTM. Our final requirement was a need for easy-to-use, easy-to-maintain tool. This custom-built tool worked great. Again, no code was required.
Building with AI
In 2024, we have seen the rise of generative AI agents. Everyone building agents is moving everyone to become software engineers. Because LLMs need information and context to learn from, references such as databases with the information are required. Introducing GitHub, your repository. Next, you need an LLM to power your agent and possibly a tool to help you build workflows. Additionally, your agent needs to consistently run and learn to enter the Mac mini (your on-premises server). So, the long and short of it is, you become a software engineer. There is a learning curve, a lot of false starts, and let’s not forget security that must be in place for the tool.
Once you have mastered your new computer engineering skills and your agents are active, your job will change again. Now you must focus on vector database maintenance and model fine-tuning. Models change; new data is created by processes and needs to be stored. Someone must pick this work up.
Buying vs AI building
With over 15,000 marketing technology tools currently in the marketplace. Everything you can think of is available to purchase. Yet there is a drive for marketers to build their own products with AI. Referring to the Subscription vs. Cost section, there isn’t much of a saving there. Consider additional cost such as lost time for your staff spent building their agents. Additionally, administering and governing this new tool is needed. When you’re purchasing the software from a third party, they maintain it, fix bugs, etc. Now this is all on you and your team. Another potential issue with building the AI app vs purchasing is if the individual who built the agent left, did they document their build? Can your other team members access the tool to administer it?
To see more information about the financial breakdown of building software in-house vs buying, check out Aaskash Gupta’s article.
Before you run to have the marketing team build their agents, realize the risks, costs, and challenges of the approach. Do you want your marketers spending time being software engineers, or do you want them writing copy, planning events, or sending marketing messages?
Where to go from here
While managing marketing technology today is overwhelming, long-term planning is still possible. LLM are based on data and context. But before we jump into the data, identify what you want the AI agent to help you with. There is so much data you could focus on cleaning, centralizing, and contextualizing. Before you boil the ocean, stop and focus on what will have the greatest impact. Is it reporting out campaign successes? Or is it helping to qualify leads?
Once you have narrowed your focus, it is time to address the data. Where is your marketing data? What state is your marketing data in? Is your marketing data complete? Have you documented your marketing processes? Are those processes stored in a place that can be accessed? Do you have a structured history of your interactions with marketing requests? Is there standard reporting? Is your report’s history in a centralized place? Start mapping it out: identify where the information is, how complete it is, and where you will need to provide cleanup and context. Document as you go along, allowing you to refer back to decisions made along the way.
Similar to building any app, you will need to oversee the agent. Each model will be updated; things can break along the way.