This blog was adapted from the AdAge article, “How marketers can evaluate AI tools – experts weigh in on where to invest” by Lindsay Rittenhouse.
Marketing leaders are being presented with a growing number of AI tools, agents and agency-built solutions, all promising greater speed, efficiency and productivity.
But how do you know which are genuinely worth the investment?
Our Founder and Co-CEO, Simon Francis, recently spoke to Ad Age about how marketers can evaluate AI investments, sharing the seven-step framework we use at Flock.
Here are four things we think marketers should consider before committing budget.
1. Start with the need, not the technology
Before evaluating an AI solution, establish whether there is a genuine problem for it to solve.
We’ve seen proposed agents duplicate capabilities clients already have, including a daily update agent replicating information provided by an account director, and a reporting tool recreating functionality already available through Google Ads at no additional cost.
Doing nothing should always be one of the options on the table.
2. Prove the value
A compelling AI business case depends on an accurate baseline.
We recommend measuring how long a task actually takes today, then comparing it with the AI-assisted version, including time, token usage and associated costs.
We’ve seen agency estimates significantly overstate the cost of existing manual processes. We’ve also seen supposedly automated agents supported by manual workarounds behind the scenes.
As Simon told Ad Age:
“What you often find [is] the agents don’t actually work. People are doing manual workarounds in the background.”
Test the solution against the real baseline before you scale it.
3. Scrutinise the cost, and who pays it
The headline build price rarely tells the whole story. Marketers need to consider total cost of ownership, including implementation, model usage, maintenance and upgrades.
Independent benchmarking can also expose significant differences between a proposed price and the likely cost of building and running an agent. We use tools that help marketers independently assess these costs, including calculators that estimate what an agent should cost to build and run. As Simon explained to Ad Age, if an agency proposes a $50,000 build, independent analysis could show that the same solution should cost closer to $14,000. That gives marketers a tangible benchmark against which to challenge the proposed investment, rather than taking the quoted price at face value.
There’s also the question of who should pay. Our view is that AI used to improve an agency’s own backend efficiency should generally be included within agency overheads, rather than charged to the client as an additional capability.
For technology that creates additional client value, different models may be appropriate, including asset-based pricing, SaaS licensing, build fees or performance-linked compensation.
4. Get the governance right
AI investment isn’t only a technology and cost decision.
Marketers need clarity on who owns the agent, its code, prompts, data and outputs, what happens if the relationship ends, and where liability sits if something goes wrong.
Contracts should also address areas including data retention, model training, security, IP and copyright, human accountability, audit logs, disclosure, portability and ongoing performance monitoring.
These questions need to be part of the evaluation process, not an afterthought once the technology has been selected.
Our seven-step framework for evaluating AI investment
The framework Simon shared with Ad Age provides a practical process for assessing AI tools, agents and agency capabilities:
|
Step 1
Define the need
|
Define the need for each use case, including required outputs, ownership, integration requirements, volume and scale, budget and success criteria. |
|
Step 2
Set the baseline
|
Record the current pre-AI process, quality, delivery time and full cost. Establish how future cost avoidance and effectiveness gains will be measured. |
|
Step 3
Identify the options
|
Identify the tools and providers that could meet the use case, not only the incumbent agency’s offer. |
|
Step 4
Screen and shortlist
|
Gather proposals, apply minimum security, data, IP and technical requirements, then select a small number to test. |
|
Step 5
Run a proof-of-value test
|
Use real briefs, brand rules, data and approval processes, and compare each solution against the baseline. |
|
Step 6
Choose the commercial model
|
Compare total cost of ownership and determine the right pricing model, including agency overheads, asset-based pricing, a transferable build fee, SaaS licensing or performance-linked compensation. |
|
Step 7
Contract and monitor
|
Agree permitted uses, data retention, security, IP, accountability, liability, ownership, portability, termination and ongoing performance monitoring. |
The bottom line
The question for marketers isn’t how quickly they can adopt more AI. It’s whether each investment solves a genuine need, creates measurable value and stands up to commercial scrutiny.
Define the need. Establish the baseline. Test the value. Understand the cost. Then invest.
That discipline will become increasingly important as AI reshapes marketing operating models, agency scopes and remuneration.
If you’d like our help to get started, fill in the form below.
Contact us