By Tristan Harrison, MD, BearJam
Each enterprise experimenting with AI-powered automation ultimately asks the identical query: how far can we take this? Get it improper in a single path and you progress slowly, whereas rivals don’t. Get it improper within the different and also you scale issues as quick as you scale output.
At BearJam, an AI-powered video manufacturing firm, we’ve spent the previous two years discovering the purpose in between. And right here’s what we’ve discovered.
Begin by deciding what needs to be automated, not whether or not it may be
We selectively use AI throughout the manufacturing course of, together with storyboarding, script growth and chosen voiceovers. Whereas some tasks can now be delivered nearly completely utilizing AI, we’ve discovered it typically works finest when mixed with live-action taking pictures, conventional post-production and human inventive judgement.
The query that issues isn’t merely “can AI do that?” It’s “the place on this workflow will AI genuinely make the work higher?”
Extremely particular branded belongings, human performances and culturally delicate particulars are normally nonetheless higher dealt with by individuals.
AI lawyer Kelsey Farish explains an vital possession facet: “Legal responsibility for AI (automation) output largely sits with whoever deploys it, not the platform that constructed the device, so manufacturers can’t outsource the pondering to their AI vendor’s phrases and situations.
“As an alternative, get forward of it, determine, and say out loud, how and why you’re utilizing AI, slightly than hoping no one asks.”
Construct human checkpoints into the method, not round it
Accountable automation doesn’t imply shifting quick and checking for issues on the finish. On our tasks, purchasers approve static AI-generated frames earlier than we transfer into animation.
Which means inventive, model and accuracy points get caught whereas they’re nonetheless simpler to repair, slightly than after a full sequence has been rendered.
The precept generalises past AI video production: put oversight at outlined factors contained in the workflow, not as a last sign-off bolted on afterwards.
Companies needs to be fascinated about the required approvals, audit trails and accountability when automation is concerned and baking that into their processes uniformly.
Automation scales your errors as quick as your output
Extra ideas, extra variations, extra belongings, in much less time: sure, that’s the upfront pitch of automation, and it’s considerably true. However the identical pace that multiplies your output additionally multiplies errors: a improper emblem, an inaccurate element, a cultural misstep, reproduced throughout each model earlier than anybody notices.
Our expertise is simple: the extra particular the output must be, the extra vital human evaluation turns into. High quality management has to scale alongside manufacturing quantity and definitely not lag behind it.
Hybrid isn’t a compromise, it’s the mannequin
Our strategy is intentionally hybrid. AI accelerates the repetitive, time-intensive elements of manufacturing; individuals retain duty for judgement, enhancing, craft and high quality management.
That break up can also be altering what “human experience” seems to be like. We employed Brick Ng as an AI Artist for instance, a job that didn’t exist earlier than demand for AI manufacturing created it.
His background in graphic design lets him translate lighting, composition, lens selection and visible style into AI workflows. Accountable automation isn’t only a course of query, it’s additionally an funding within the individuals who function, oversee and optimise the technological output.
Decide the entire workflow, not the automated step
Automating one stage doesn’t routinely make the entire course of sooner or cheaper. Constant characters, exact human performances and sophisticated scenes can nonetheless want heavy iteration and post-production, and in some circumstances conventional filming is easier and no dearer.
The objective needs to be a greater total course of, not the best potential variety of automated duties. Which means being prepared to change between automated, conventional and hybrid approaches relying on the temporary, slightly than defaulting to whichever one is trendy.
Clear limits make it simpler to innovate, not tougher
Figuring out the place you received’t use the expertise is as a lot part of accountable automation as realizing the place you’ll. We’ve turned down tasks we may technically ship, together with deceptive before-and-after product imagery.
Having that line drawn prematurely means the staff can experiment freely inside it, as a substitute of re-litigating the identical threat each time a brand new temporary is available in.
Most significantly, companies may have firmer crimson strains than “gray space however technically high-quality” as a result of the authorized or reputational draw back outweighs the effectivity achieve.
The objective is better-designed automation, no more of it
None of this requires slowing each course of down with additional layers of sign-off. It means automating the suitable duties, retaining human judgement the place it provides worth, and constructing checks into the workflow from day one slightly than including them after one thing goes improper.
Our expertise suggests the strongest mannequin isn’t human versus AI. It’s a hybrid course of that makes use of automation for pace and scale with out designing experience out of the system. Companies can transfer shortly with automation, so long as duty is constructed into how the workflow operates, not bolted on after the actual fact.
About BearJam
London’s award-winning video manufacturing company. Daring content material for Netflix, Purple Bull, Revolut and extra. Company video, AI video, animation.
BearJam is the primary hybrid AI UK video manufacturing firm to commit 1% of revenue from AI-intensive projects to nature restoration, supporting the long-term restoration of a 69-acre habitat website in Kent.
