Vol. III · Essays on work after AI
Saturday, October 3, 2026
evolving.work
Knowledge · Essay № 054

Why Are We Still Communicating Like Video Is Expensive?

The cost of producing visual explanations has collapsed by 90-99% with tools like Sora, Veo, and Synthesia.

H Héctor Reinoso Aug 16, 2026 13 minute read Share · Save

Consider a slide deck that explains a visual process in forty-seven bullet points. It exists because someone, three years ago, decided producing a four-minute video walkthrough would cost more time and political capital than another deck. They were right at the time. The deck got built. The deck gets opened in onboarding. The deck has been “refreshed” twice.

The cost calculation that produced that deck no longer holds. We are still living inside its decisions.

The cost of producing a clear visual explanation has collapsed structurally, and most organizations have not noticed. We are running a 2023 communication budget against a 2026 production economy. The gap shows up as slides where videos would be clearer, meetings where async demos would suffice, and L&D teams spending a month refreshing PDFs because video felt expensive when the org chart was drawn.

The Constraint That Shaped How We Communicate

Producing video used to require coordination. Script. Filming. Editing. Lighting. Often an external producer for anything that touched a customer or a CEO. Even internally, anything past a screen recording demanded a budget line, a calendar block, and a small number of people whose week was about to disappear.

That coordination cost did real work. It acted as a filter. If a topic required a video, it had to matter enough to justify the spend. Compliance refreshes mattered. Quarterly all-hands mattered. Process explanations for a single team usually did not — so they went into slides. Cross-team handoffs did not — so they went into a meeting.

The medium got chosen on cost, not clarity.

That format choice rarely gets audited. The format of an organization’s communication is seldom a deliberate design decision. It is a fossil record of what was cheap to produce in the years when the workflow first calcified — built when 61% of the U.S. workforce was not yet remote or hybrid [9].

A 2025 Canva and Harris Poll study used Steady State Topography on 63 professionals and found creative visual content triggered memory encoding 74% faster than dull alternatives, with 21% greater emotional intensity. The slide deck is not just less engaging. It is less memorable, and we have known it for years — routing around the finding because video was expensive, the way you route around a leak in the roof until you stop noticing the bucket.

That Constraint Is Gone

The numbers carry the argument. Traditional internal video production runs $1,000 to $10,000 per finished minute. AI-generated video runs $0.50 to $30 [1] — a 90-99% reduction in unit cost. A 10-video corporate campaign that would have cost $100,000 or more through a traditional agency now costs roughly $89 through AI tools [12]. That comparison shows how far the quoted production costs have fallen.

The National Bureau of Economic Research now reports the unit cost of producing video using tools like Sora and Veo is lower than the unit cost of producing the same content using ordinary workers [2]. That is the economic threshold at which the cheaper format becomes the rational default.

Markets are reacting. The AI video generation market sat at $788M in 2025 and is projected to reach $3.44B by 2033; enterprise spending grew 127% year over year [5] — finance teams signing six-figure renewals for a category that did not exist in budgets three years ago.

This is the first communication-cost collapse since email. The pattern will look familiar, and the lesson we mostly failed to learn the first time is still sitting in view.

But the System Has Not Adjusted

Video has an 85% effectiveness rating for internal communications. Only 43% of organizations use it as a primary channel [3]. The difference between those numbers is not a quality problem. It is a habit problem.

77% of internal communicators now use AI tools for video. 60% still rate their engagement outcomes as average [4]. The tools are in the building. The workflow is still the workflow designed for slides. Buying a Sora subscription and routing the output through the same approval chain that took six weeks for a PDF is a particularly modern way to waste $200 a month.

Some will object that cost was never the real bottleneck — workflow design is. They are right, and it is the deeper point. Cost collapse is necessary but not sufficient. When the cost barrier disappears, the structural obstacles — workflow inertia, missing defaults, no decision framework for format selection — become visible for the first time. Most organizations were using “video is expensive” as a stand-in for “we have not built the workflow for video.” Lower production costs leave the approval chain and the workflow around it open to examination.

Research on organizational inertia confirms it: communication patterns persist long after the original constraints have changed [6]. And studies of remote leadership find videoconferencing quality is the only channel that uniquely predicts engagement among distributed employees. With 61% of the workforce now remote or hybrid [9], async visual communication is the channel best matched to the conditions of work — a polite way of saying the deck-based workflow was designed for a building most of us no longer go into.

The Harris Poll’s 2025 Visual Economy report estimates organizations spend $143 billion annually on visual content while creative teams juggle eight tools weekly and produce 25 hours of work that could take 10. The pain is not a lack of technology. It is a 2018 workflow inside a 2026 toolchain.

What Changes When Production Becomes Cheap

The argument so far is structural. Three concrete cases make it specific.

Duolingo. With 100M+ monthly active users, Duolingo faced a structural problem: visual explanations of grammar and culture were prohibitively expensive across 40+ languages. Traditional video ran $3,000-$8,000 per finished minute, so visual explanations existed only for the top five. Everyone else got static text — not because text was better, but because video was unaffordable [10]. In 2024-2025, Duolingo integrated AI content tools. The format-selection question changed from “can we afford video for this language?” to “does this concept benefit from visual explanation?” Coverage expanded to lower-resource languages that had been text-only by default.

Cost collapse did not just make existing videos cheaper. It unlocked formats that were uneconomical before.

Synthesia enterprise clients. Xerox replaced slide-based onboarding with AI video walkthroughs. BSH used Synthesia to scale multilingual training across 30+ countries without per-language production costs. Per-minute cost dropped to roughly $2.13 [1]. Quarterly content updates — previously untenable — became routine. The lesson is sharper than “AI video is cheap.” Organizations default to static formats not because static is more effective, but because dynamic formats were too expensive to maintain at the cadence the business moved.

The async-video pattern. This vignette is a composite based on patterns across professional services firms. A 200-person consulting firm spanning three time zones averaged 12 meeting-hours per consultant per week. Knowledge sharing and handoffs defaulted to live meetings because producing a recorded walkthrough cost more than scheduling another meeting. A new Chief of Staff implemented a “video-first” policy: any 30-minute meeting should first be attempted as a 5-minute async video. Meeting hours dropped from 12 to 8 per consultant per week. The library accumulated 200+ assets in three months.

Meetings persist for the same reason slides do: the alternative — producing a reusable visual explanation — was historically more expensive than gathering people in a room. That cost just inverted.

Other training programs show the scale of the potential savings. IBM saved $579 million over two years after shifting 50% of training to video. Microsoft cut per-person classroom training from $320 to $17. 47% of L&D teams were implementing microlearning in 2024. The market shift is not a forecast. It is a pattern named organizations have already paid for.

More Video Might Mean More Noise

Two objections the thesis cannot wave away.

The first is quality. AI video is not yet at a level where it should replace high-trust formats — executive communications, regulated compliance training, client-facing brand work. If an enterprise sends uncanny-valley training videos under a recognizable logo, the brand cost may exceed the savings. Deloitte’s 2026 TMT predictions describe generative video as “approaching Hollywood quality” while flagging risks of overwhelming audiences and eroding authenticity [8]. The thesis is not that every communication should become video; it is that most internal communications never justified professional production in the first place. The 85/43 gap is about formats that should already be video, not about whether your earnings call should be a Sora clip.

A 2025 Journal of Consumer Research study by Min, Liu, and Anderson found consumers infer production cost from visual complexity — simple aesthetics get associated with lower production effort, which lowers willingness to pay. Polish still carries signal. The cost of producing the right visual explanation for the right audience has fallen by an order of magnitude — not the rules for choosing one.

The second objection is volume. When production costs collapse, output expands. If every manager can now produce a video instead of sending an email, the result may be more noise — and video makes overload worse before it makes it better, being harder to skim, search, and cite.

We have run this experiment before. The cost of sending email approached zero, and volume exploded. Video production costs just made the same move. Organizations that do not build curation systems before they scale generation will recreate the email overload problem in a slower medium. The argument for making intelligence cheap was always that scarcity moves; this is communication’s version of the same lesson.

And “cheap to generate” is not “cheap to deploy at scale.” A Sora clip still costs roughly $1.30 in compute per 10 seconds, and Forbes reported OpenAI burning $15 million per day to support Sora. The cost curve has fallen unevenly: workflow integration, governance, rejected drafts, and review cycles all add friction the marketing leaves out.

Visual explanation has gotten cheaper. It has not become free. Organizations that profit from the shift will treat curation as the new scarcity.

The Communication Format Audit

To examine which formats have persisted out of habit, work through five steps.

  1. Map current format defaults. For every recurring communication type — onboarding, training, status updates, handoffs — list the format used, the reason it was chosen, and whether it matches the content. Most teams skip this because they assume their current formats are deliberate. They almost never are. Common failure: treating inherited defaults as designed defaults.

  2. Identify format mismatches. For each flow, ask three questions. Is this concept inherently visual? Is this explanation repeated for different audiences? Would this content benefit from being searchable and reusable? If yes-and-still-static for more than one, flag it. Common failure: prioritizing by production difficulty rather than communication impact.

  3. Calculate the cost inversion. Compare the current cost of the format — participant time, repetition, comprehension failures — against the new cost of the better format. For a meeting, multiply participants by hourly rate by duration by cadence. For a slide deck, include creation time plus comprehension time lost across the audience. Test on three high-frequency flows. Common failure: counting only production cost of the new format and not the hidden cost of the current one.

  4. Redesign the workflow. Pilot a “video-first” default for two or three high-impact flows. Pick the production tool. Define storage and search. Set an update cadence. Measure adoption and comprehension. Common failure: adopting the tool without changing the workflow — producing AI videos and distributing them through the same channels, with the same governance, as the documents they replaced.

  5. Govern volume and quality. Before producing a video, ask whether the content is visual in nature, will be consumed by more than five people, and will remain relevant for more than two weeks. If fewer than two of three are met, text remains the default. Review the library quarterly. Common failure: skipping this step, which is how email overload happened.

McKinsey calls AI’s underlying shift “the democratization of production itself” [7]. The same applies to communication. The work has shifted from making things to deciding what is worth making [11] — a harder question than the cost question, and the one the cost collapse has finally made unavoidable.

We spent fifteen years building organizations around the cost of producing a clear explanation. That cost just fell by a factor of a thousand. The remaining question is not whether we can afford to explain things well. It is whether we are still choosing not to.

References

Supplementary sources (enrichment):

AI strategyAI transformation
About the author
Héctor Reinoso writes about how organizations should be designed when intelligence becomes abundant. He works on enterprise AI adoption inside a large industrial organization, after 25 years as a strategy consultant and transformation executive across four continents. Harvard MBA, industrial engineer, and still writes the code he writes about. More →
Continue reading
Knowledge · Aug 19, 2026

Is YouTube Still Learning — Or Just Repetition?

When content production cost collapses, volume rises but novelty compresses.
Knowledge · Aug 14, 2026

Stop Ruining Email

AI has driven the marginal cost of sending email to near zero, triggering a self-reinforcing arms race: volume-driven outreach floods the channel, aggressive filtering catches…
Org Design · Jul 25, 2026

Why AI Pilots Fail (And Why You Should Run Them Anyway)

AI pilots can build organizational skill even when they produce no immediate return. The test is what the organization learns and retains.
Knowledge · Jul 5, 2026

Long Context Is Not the Same as Better Memory

Long context and memory solve different problems; confusing them is the most common error in AI system design
Knowledge · Jun 23, 2026

Curate Before You Automate

Treat curated external knowledge as owned infrastructure; curate the reference layer before you automate against it
Knowledge · Jun 8, 2026

Don't Build Your AI Strategy on Someone Else's Stack

A framework for AI infrastructure risk, grounded in what frontier companies themselves disclose about dependency