AI training gets a person to competent. What follows determines whether that competence becomes firm capability.
The eight-session AI marketing and business development course I am teaching right now has two separate certification requirements: attend at least six of the eight live sessions and complete at least six of the eight homework assignments. I track and score those two requirements apart from each other, on purpose.
The assignments are not academic exercises. Students repurpose a real client alert into a LinkedIn post, an email summary, and a set of social posts. They draft an RFP response, a Chambers or directory submission, a prospect research briefing built from two different AI tools, a ninety-day competitive scan of one rival firm's lateral hires and new practice launches, a one-page firm AI style guide. The subject matter is whatever is already sitting on their desk that week.
Four sessions into the current cohort, that split between attendance and practice is already visible in the data.
One student has attended nearly every live class and submitted none of the three homework assignments due before today's session. Their attendance record is fine. Their practice record is empty.
Two other students each missed a live class entirely and still completed that week's assignment on their own. They did not sit through the session, but they used the material on real work.
A business development director's calendar does not hold open blocks reserved for homework. A client call runs long, a partner needs something turned around by five, and the assignment moves to next week, then the week after that.
The distinction is not a judgment about effort or ability. Missing a class does not mean someone lacks commitment. Attending without submitting an assignment does not mean they learned nothing. The two records answer different questions.
Attendance tells me whether someone showed up to learn a skill. Homework tells me whether they used it after the instruction stopped.
Law firms conflate those two things. Someone attends a workshop, finishes a course, or earns access to a tool, and the firm records that person as trained. That record is accurate as far as it goes. It says nothing about whether the skill has become part of how the person actually works.
I built this course because people need instruction, examples, feedback, and a place to ask questions while they are learning. A class can cut the time it takes to go from a blank prompt box to competent use. It cannot supply the repetitions that follow.
If I want to know whether the training will change how someone works six months from now, the practice record tells me more than the attendance sheet. That happens through the repeated, unglamorous work of using the tool on real assignments, with no instructor there to catch a mistake.
Training gets a person to competent, and consistent practice makes that competence durable.
Durability has a ceiling, though. A marketing or business development professional can build the habit and still have no voice in the firm's decisions about AI.
The assignments also show why a single exposure is not enough.
Across the current cohort, I have seen the same category of mistake return in different forms from one week to the next, even after we addressed it in class. A student repurposes a client alert into a finished LinkedIn post but leaves out the prompt that produced it. The output may be good. Without the prompt, no one, including the student, can tell whether the result is repeatable.
Another assignment asks for a firm AI style guide. The submission explains the tone the firm wants but says nothing about what the model should avoid. That produces a document a person can read and agree with, but an AI model has no concrete boundary to check against.
These are small craft habits, and neither one is hard to grasp once it is explained live. Keep the prompt with the output. Give the model specific exclusions instead of vague taste.
Grasping a rule in a room is not the same as remembering it three weeks later, at nine at night, with a real deadline attached. That gap between learning it in class and using it three weeks later under deadline is not a discipline problem. It is what learning a new tool under a real workload looks like.
The lesson starts to stick when a student includes the prompt on the next assignment without being reminded. The style guide becomes useful when the student starts writing exclusions because they have learned to anticipate what the model needs. The correction has become part of how they work.
Good instruction gives people a sound method and feedback while they apply it. Practice carries that method into the next draft, then into the regular flow of client work. Without the second part, competence stays dependent on the classroom.
That dependence is easy to miss because attendance is cleaner to track. There is a calendar invitation, a participant list, and a completion record. Practice is messier. Someone has to review the work and determine whether the method is improving. It takes more attention than counting names on a screen, because it means reading what a student produced this week against what they produced the week before and deciding whether the correction held.
The separate requirements in my course force both questions to stay visible. Did the student attend enough instruction to learn the method? Did they complete enough applied work to begin forming the habit?
A certificate should answer both. A firm's capability requires one more question after that: does the person who built this skill have a formal voice in how the firm decides to use AI at all?
Every firm I train runs its AI decisions through some kind of council or steering committee, and in nearly all of them, no one from marketing or business development has a seat on it.
I told the current cohort: "I can't tell you how many firms I talk to that have an AI council with no one from marketing or business development on it. To me, that is a big mistake, because no one knows your clients better than your marketing and business development team."
That exclusion separates two forms of knowledge the firm needs in the same room. The council has responsibility for decisions about AI across the firm. Marketing and business development bring the closest understanding of the client.
Asking for that seat is not a small thing to ask, especially for a department that has never had a standing one. It means walking into a room that has always belonged to IT and the general counsel's office and making the case that client judgment belongs there too, usually without the title or the tenure that makes that case easy to make.
When that perspective is missing, the people building useful AI habits in marketing can improve their own work, but their judgment stays confined to their own work. They can produce a more repeatable draft or a more useful style guide. They cannot carry what they are learning into the forum that determines how AI will be used across the firm.
This is why the training question and the governance question belong together. Treating them as two separate problems, solved in whatever order happens to be convenient, produces one of two incomplete results.
A seat on the council without practiced competence can become ceremonial. The marketing or business development representative is present, but they do not yet have enough direct experience to translate client knowledge into useful guidance about AI.
Practiced competence without a seat creates the opposite limit. The team develops real judgment through repeated use, but the firm makes broader decisions without it.
Training has to come before the seat does. Training and practice create an informed contributor, and formal inclusion gives that judgment a vote in the room.
Marketing and business development should not be brought in after an AI decision has been made so they can explain it to everyone else. They should be part of the group making the decision.
For CMOs and marketing or business development directors sponsoring AI training, none of what follows requires a training budget you do not have or a headcount you were never given.
Track attendance and applied work as separate measures. A participant list is not evidence of adoption on its own. Ask which real assignments people applied it to. Read the prompts next to the outputs so repeatability can be checked instead of taken on faith. When someone builds an AI style guide, look for specific instructions about what the model should avoid; a description of the desired tone alone will not do the job.
Those artifacts reveal more than a completion badge can. They show whether a person can reproduce a result and whether corrections are carrying into the next assignment. They also give the team concrete experience to bring into the AI council conversation.
Then connect that practice to governance. Marketing and business development need standing representation on the AI council or steering committee, with the same opportunity to help make decisions as the other functions involved. The person in that seat does not need to be the most technical person in the firm. They need enough applied experience to know what the tools can do, where human judgment remains essential, and how the firm's choices will affect the client.
That connection changes what the training produces. The individual's practice is no longer just a task getting done at their own desk. It becomes evidence the firm can point to when it decides who sits on the AI council and what that person is allowed to weigh in on.
I also designed this course with two separate bars because I do not want attendance confused with application. Six sessions show that someone was present for the instruction. Six assignments show that they kept working after each session ended.