Traffic and scheduling is one of the least glamorous functions at a TV station and one of the most important. If the traffic log is wrong, the revenue is wrong. If the schedule has errors, the viewer experience suffers and advertisers get unhappy. Traffic managers carry a lot of weight at LPTV stations, and in many small operations the function is being handled by someone who is also doing two or three other jobs.
So when I talk to operators about where AI might genuinely help, traffic and scheduling comes up early. The work is rule-based enough that automation can handle meaningful portions of it, it is detail-intensive enough that human errors happen regularly, and the consequences of those errors are direct and measurable. This is a good match for what AI tools are currently capable of. Let me walk through what is actually working now, and what I would encourage stations to approach carefully.
What traffic and scheduling actually involves
Before getting into AI, it is worth naming what the function involves. At most LPTV stations, traffic and scheduling covers a cluster of tasks. Spot placement against sold inventory. Log building. Conflict resolution when placements overlap or violate advertiser rules. Make-good management when spots miss or run in wrong positions. Program scheduling. Spot substitution when a scheduled spot is pulled. Reconciliation between what was scheduled and what actually aired. Invoice preparation based on airing proof.
Some of these tasks have been partially automated for years through traffic software. What is new is that AI tools can now handle or assist with more of the judgment-heavy portions of the work.
Where AI is actually helping today
The most common application I see working is log optimization. Given a set of sold spots with placement requirements, advertiser separation rules, and inventory constraints, AI can propose log arrangements that better balance competing demands than a human building the log manually. This is not a dramatic change, but it saves time and catches conflicts that might otherwise slip through.
A second application is conflict detection. AI tools can scan a proposed log and flag situations where two competing advertisers are too close together, where an advertiser’s placement requirements have been violated, where program content doesn’t match advertising content in ways that could be problematic, or where legal requirements have not been met. The tools are not perfect. A human still needs to review the flags. But they catch things a tired human eye would miss at 5 p.m. on a Friday.
A third application is make-good optimization. When a spot misses or runs in a wrong position, figuring out the right make-good placement requires balancing the affected advertiser’s needs, available inventory, and other constraints. AI tools can propose options that a traffic manager can review and choose among, which is faster than working through the options manually.
A fourth application, which is still maturing, is predictive scheduling. Given historical audience data, programming information, and known events, AI tools can suggest program scheduling changes that are likely to perform better. The accuracy of these predictions varies considerably, and I would not rely on them as the sole input to a scheduling decision, but they can be useful as one perspective among several.
The integration question
One complication worth naming is that many AI applications in traffic and scheduling require integration with the station’s traffic software. The major traffic systems are at different stages of incorporating AI capabilities directly into their platforms. Some are further along than others. What this means practically is that the AI capabilities available to your station depend significantly on which traffic system you use and which version you are running.
For stations on older systems, there are workarounds. You can export data, run it through AI tools externally, and use the results to inform decisions in your traffic system. This is clumsy but workable. For stations considering traffic system upgrades, AI capability is now a relevant factor in the selection.
What to watch out for
A few cautions based on conversations with stations that have tried various AI traffic applications.
Accuracy degrades with bad input. If your sold inventory data, advertiser rules, or program information is messy or out of date, AI tools will produce bad recommendations based on that bad data. They will not know the data is bad. Cleaning up your underlying information is often the highest-value first step before adding AI to the process.
Edge cases still require human judgment. Unusual situations. Political advertising rules. Specific advertiser relationships that aren’t captured in the rules database. The more context a situation requires, the less reliable AI recommendations will be. Your experienced traffic manager’s judgment remains valuable precisely for these cases.
Audit trail matters. If a spot runs in the wrong position and an advertiser complains, you need to be able to reconstruct why it ran where it ran. AI tools that produce recommendations should be used in ways that preserve a clear record of what was recommended, what was accepted, what was overridden, and by whom. This is a policy and workflow discipline, not a feature of the tools themselves.
A realistic picture of the productivity gain
Stations that have adopted AI tools thoughtfully in traffic and scheduling generally report moderate but meaningful time savings. A traffic manager who used to spend six hours a day on log-related tasks might get that down to four. The two freed hours can be redeployed to higher-value work, whether that is more careful conflict review, better reporting, or assisting with other station functions.
What I would not expect is dramatic headcount reduction. The work is still detail-intensive, still requires judgment, and still benefits from an experienced human at the center of the process. Stations that try to replace traffic managers with AI tools are going to create problems for themselves that will show up in advertiser complaints, revenue errors, and operational disruption.
A suggested starting point
If you are interested in exploring AI in your traffic operation, I would start with two questions. What are the two or three most common sources of error or friction in your current workflow? What AI capabilities, either within your traffic software or as external tools, could address those specific issues?
Starting with specific problems rather than general enthusiasm tends to produce better results. You can measure whether the problems got better. You can decide whether the investment was worth it. You can expand to additional applications if the first ones work.
This is how almost every other operational improvement in broadcasting has happened. Identify a specific pain point. Try a specific solution. Measure the result. Expand what works. AI is not different in this respect. It is a tool to solve specific problems, and the stations that will get the most out of it are the ones who use it that way.
