<- all tokdocs
2026.09.13@sourcetms

The AI Rollout Workshop in This TikTok Is Gated Behind a Lead Form, and Already Free on YouTube

AI
Watch on TikTok

View on TikTok ->

The landing page this video walks through asks for a first name, phone number, and company email to unlock a workshop recording that has been public on YouTube since September 9, 2026. The 50-second clip is a talking-head video over a screen recording of specialtytokens.com/executive-ai-workshop. Thomas Schlossmacher says he and his co-founder were invited by "the Federal Chamber of Commerce" to present an AI rollout to executives, that a few attendees called themselves dinosaurs, and that commenting the word "link" gets you the recording and the slides. On screen, the page header reads "From AI hype to company capability," with a nav bar of Services, Client results, How we work, About, Contact, a DE language toggle, and a "Book a free audit" button in the top right. An announcement strip at the very top reads "NEW · Next remote executive AI session · Register interest →."

What the screen recording actually contains

The body copy under the hero reads: "The executive AI workshop we run in partnership with the Austrian Federal Economic Chamber (WKO), for owners, managing directors and the heads of finance, operations and sales. Unlock the full recording and the slides with your first name, phone number and company email." Next to it sits a red "Watch and open the slides →" button and the red-and-white WKO logo.

Below that is an embedded YouTube player titled "Executive Session - How To Run an AI Roll Out," from the channel "@sourcetms - AI Builds & Signals | Source AI," with a total runtime of 36:55 in the corner. Schlossmacher scrubs through it during the clip, landing on 2:05, 3:55, 4:14, 16:42, and 21:11. A sidebar card reads "The slides. The complete deck, in the browser, ready to share with your leadership team," with an "Open the slides →" button.

Further down are two photos of the room, captioned "In the room: leaders from companies with 500 million in revenue down to 30-person high-performing teams, from manufacturing to services." The slides visible on the conference room screens read "Every agent wakes up knowing the world, and nothing about your company," with the words "your" and "company" highlighted in blue. Other slides that flash past in the player read "A chatbot is not an AI strategy," "AI is part of every headcount request," "The lead's job is a loop," and "The employee does the work. AI helps." At 4:14 the player shows a clip with a lower third reading "Peter Steinberger," the Austrian developer who founded PSPDFKit. The last section visible before the video cuts is a heading: "Four parts, from what changed."

The gate and the free version

I checked the live page. The four-part curriculum, the WKO copy, and the "In the room" photos are all publicly readable at specialtytokens.com/executive-ai-workshop. The recording and the deck are not. Both sit behind buttons labeled "Unlock the slides and video →," which is the lead form the on-screen copy describes.

The recording itself is a public YouTube video. Video ID 3bZ91-3T_OU, published 2026-09-09 at 00:17 Pacific, 2,215 seconds long, which is exactly the 36:55 shown in the embed. YouTube's player data lists it as public, not unlisted, and it appears in the channel's regular video listing. When I pulled the channel page on 2026-09-14 it showed 231 views, and the watch page showed 232. The YouTube channel handle is @sourceaishow, not @sourcetms, which is only the display name prefix. Anyone who searches the exact title finds the whole session without giving up a phone number.

The slide deck is a different story. specialtytokens.com/slides/s/ai-capability-workshop, the link in the YouTube description, returns a page titled "Specialty Tokens - Corporate AI Training" with no deck content served to an anonymous fetch. That one appears to be genuinely gated.

There is also a separate path at specialtytokens.com/wko, which redirects to /wko/access and asks for a password, not a lead form. That page is branded "WKO Wien," the Vienna regional chamber, and says "This WKO Wien page is private" above an "Open briefing" button. The public workshop page and the video both describe the partner as the national Austrian Federal Economic Chamber.

The four parts, in the workshop's own words

The section the video cuts away from is fully readable on the live page. The workshop is structured as:

The reality check covers "what really changed this year, and the four levels a company sits on: no company AI, company-wide chat, connected internal systems, or work that runs as an agent workflow." Build your first AI skill is the hands-on portion, where "each executive takes a task they already do well, captures it as a reusable skill, and saves it where the company can reach it." The company-wide picture covers "one central AI platform, agents at every seat, meeting memory, and what happens to the things people build once the workshop is over." Your company roadmap covers "three ways in, the point person's actual job, governance and adoption, and what months one, three and six look like in practice."

The four-level maturity ladder is the strongest idea here, because it gives a leadership team a concrete place to stand. A company that has only bought ChatGPT seats has a different next move than one that has already wired an agent into its ERP. The weakest part is the second segment. Having each executive capture one task as a reusable skill in a two-hour session produces artifacts that usually die the week after, which the page half-acknowledges by asking "what happens to the things people build once the workshop is over."

Checking the numbers the deck uses

The YouTube description lists the specific claims the session makes. Most of them check out against primary sources, with two caveats worth naming.

The METR figure is slightly off. The description says "4h 53min human-length tasks at 50% success." METR's published estimate for Claude Opus 4.5 is 4 hours 49 minutes, with a 95% confidence interval running from 1 hour 49 minutes to 20 hours 25 minutes. METR's later Time Horizon 1.1 methodology puts Opus 4.5 at 320 minutes, about 5 hours 20 minutes, with a 170 to 729 minute interval. The four minute gap does not matter. The confidence interval spanning an order of magnitude does, and no version of the slide I can see mentions it.

The McKinsey split is real. Roughly 80 percent of respondents say AI improved their individual productivity, while 37 percent say AI has contributed to their organization's EBIT, a share essentially flat year over year. McKinsey's own site blocked my automated fetch, so I verified those two numbers through secondary coverage of the survey rather than the source page.

The Ramp spending gap is real and understated. The description says AI spend per employee varies "over 600x." Ramp's Economics Lab found the top 1 percent of adopters spent a median of $7,401 per employee per month against $12 for the median adopter, which is roughly a 620x gap.

The Klarna number is accurate and incomplete. Klarna's February 2024 press release says the assistant handled 2.3 million conversations, two-thirds of all customer service chats, in its first month, equal to the work of 700 full-time agents. What the deck does not mention is that Klarna publicly walked this back in 2025 and began rehiring human agents after quality complaints. Using Klarna as a rollout success story in 2026 without that second half is the kind of thing an executive audience should push back on.

Ramp's own underwriting case is well-sourced. Ramp's blog describes card applications dropping from a two-day underwriting turnaround to as few as four hours after a two-week sprint. Flexport's numbers, roughly 2 percent lower ocean freight spend and about 20 percent better transit times, trace to public comments by CEO Ryan Petersen in late 2025 rather than a published report, so treat them as company-reported. The Shopify claim is solid. Tobi Lütke's memo, which he posted publicly in April 2025 after it leaked, requires teams to demonstrate why they cannot get what they want done using AI before requesting more headcount.

One figure I could not verify at all is "$1.3M/month model bills." I found no primary source for it, and I am not going to guess which company it refers to.

Who is behind this and why the WKO angle works

Specialty Tokens is a Vienna AI engineering firm at Kleistgasse 18, 1030 Vienna, run by two directors, Nik Redl and Thomas Schlossmacher. Redl co-founded Mokker.ai, a generative AI product photography tool acquired by Soona in March 2024. Schlossmacher runs the sourcetms brand across TikTok, YouTube, and Substack, and previously built Agentbase and a VC-backed billing platform. The firm also maintains Supercenter, which it describes as an enterprise MCP and agent platform. The "Nick" named in the TikTok transcript is Nik Redl.

The WKO tie is the interesting part of the distribution model. Membership in the Austrian Economic Chamber is mandatory under Austrian law for essentially every operating business outside agriculture, which puts hundreds of thousands of member companies in one funnel. Specialty Tokens lists Wirtschaftskammer Österreich in its client strip and publishes a case study dated 2026-09-01 describing a series of in-person executive sessions. That case study states no attendance figures, no measurable outcomes, and no client quotes. It is a logo and a description of the format.

The positioning line on the workshop page is the honest one: "Every example in the deck comes from a company we have worked in." That claim sits awkwardly next to a deck whose named examples are Flexport, Ramp, Shopify, and Klarna.

Key Takeaways

  • The recording gated behind a name, phone, and company email form at specialtytokens.com/executive-ai-workshop is public on YouTube as video 3bZ91-3T_OU, published 2026-09-09, 2,215 seconds long, with 232 views when I checked on 2026-09-14. The slide deck at /slides/s/ai-capability-workshop does appear to be genuinely gated.
  • The workshop's four-level maturity model is the reusable idea: no company AI, company-wide chat, connected internal systems, work running as an agent workflow. It gives a leadership team a position before it gives them a purchase.
  • The deck cites Claude Opus 4.5 completing "4h 53min" human tasks at 50 percent success. METR's actual published estimate is 4 hours 49 minutes with a 95 percent confidence interval of 1 hour 49 minutes to 20 hours 25 minutes, and their revised TH1.1 method gives 320 minutes with a 170 to 729 minute interval.
  • The Klarna example is cited accurately for February 2024 and omits the 2025 reversal, when Klarna rehired human support agents after quality problems.
  • Ramp's Economics Lab measured the top 1 percent of AI adopters spending a median $7,401 per employee per month against $12 for the median adopter, about 620x, which supports the deck's "over 600x" claim.
  • The video claims the invitation came from "the Federal Chamber of Commerce." The public page says the Austrian Federal Economic Chamber, while the password-gated /wko page is branded "WKO Wien," the Vienna regional chamber. I could not determine from public sources which entity actually booked the session.
  • Specialty Tokens' WKO case study, dated 2026-09-01, publishes no attendance count, no outcome metric, and no client quote.
  • I could not verify the "$1.3M/month model bills" figure against any source.

Resources

Published September 13, 2026. Writeup generated from a favorited TikTok.