MarTech outlines four AI approaches and their differing costs
Rule-based, predictive, generative, and agentic AI systems carry distinct costs and risks for marketing and sales tasks according to MarTech.
AI systems often exceed task requirements
Enterprises apply complex AI systems to tasks that do not need them, according to MarTech. The initial promises of what an AI system can do often do not account for long-term costs, the human oversight required, and the total cost of ownership. Platform buying conversations occur at the feature level while the mechanism sits below where the real costs and risks lie.
Four mechanisms ordered by complexity
Nearly every AI feature uses one of four mechanisms ordered from simplest to most elaborate: rule-based, predictive, generative, and agentic, according to MarTech. Rule-based systems apply if-this-then-that logic written by hand. They cost almost nothing to run and produce the same output every time. A predictive model scores an input against patterns in historical data and remains cheap per call. A generative model produces new text or images using a large language model from open-ended input. An agentic setup wraps a generative model in a loop with access to tools that plans a step, calls the tool, checks the result, and runs again.
As complexity increases, flexibility and reach go up while predictability and auditability fall away. Risk increases from small errors to mistakes that could cost customers.
Refund request example shows cost spread
A customer refund request illustrates the differences. A rule-based system follows a policy table for automatic approval under set conditions such as under $50 and within 30 days. A predictive model scores each request as approve, deny, or escalate. A generative model drafts a resolution based on the customer phrasing regardless of prior requests. An agentic model runs in a loop to retrieve the order record, apply the policy, issue the refund, and send confirmation.
The agentic approach covers cases the other three miss yet carries the highest price by a wide margin.
Cost trajectory and accountability phase
What was four cents in 2023 is roughly $1.20 in 2026 according to EY accounting of one customer-service interaction priced first as a simple chat and then as an orchestrated agent workflow. This represents essentially the same job at about 30 times the cost. The invoice from the model provider captures only part of actual spend. Enterprises move from AI experimentation into a phase of accountability where these differences matter.
