WEBVTT 00:00:00.000 --> 00:00:03.379 If you are listening to this right now, chances 00:00:03.379 --> 00:00:05.480 are you're a retail executive. Oh, definitely. 00:00:05.599 --> 00:00:08.539 And if that's the case, you are probably intimately 00:00:08.539 --> 00:00:13.400 familiar with a very specific recurring operational 00:00:13.400 --> 00:00:16.780 nightmare. We all know the one. Right. So a massive 00:00:16.780 --> 00:00:20.559 new capsule collection drops on a Monday. And 00:00:20.559 --> 00:00:24.070 by Wednesday, your frontline associates are expected 00:00:24.070 --> 00:00:26.230 to sell it. I mean, they need to know the hidden 00:00:26.230 --> 00:00:28.870 features, the exact fabrications, how to style 00:00:28.870 --> 00:00:30.829 it. How it compares to last season's bestseller. 00:00:30.910 --> 00:00:33.189 They need to be, you know, absolute experts. 00:00:33.490 --> 00:00:36.049 Exactly. But the high quality training video 00:00:36.049 --> 00:00:37.750 that's supposed to give them all that confidence, 00:00:38.530 --> 00:00:40.329 it isn't scheduled to shoot until next month. 00:00:40.450 --> 00:00:42.609 Yeah, it is the classic execution disconnect. 00:00:42.710 --> 00:00:44.490 I mean, the product is physically right there 00:00:44.490 --> 00:00:46.109 on the floor. The customer is walking through 00:00:46.109 --> 00:00:48.289 the door. But the critical product knowledge 00:00:48.289 --> 00:00:50.590 is just stuck somewhere in a post -production 00:00:50.590 --> 00:00:54.570 editing bay. Welcome to this deep dive. Because 00:00:54.570 --> 00:00:57.170 if you've ever felt the pain of that exact scenario, 00:00:57.710 --> 00:00:59.850 you are going to want to hear what we found in 00:00:59.850 --> 00:01:02.649 today's Intel. We are looking at a stack of sources 00:01:02.649 --> 00:01:05.590 today. We've got press releases, product briefs, 00:01:05.790 --> 00:01:07.969 a blog post from a company called Multimedia 00:01:07.969 --> 00:01:11.670 Plus, or MMP. And they recently launched this 00:01:11.670 --> 00:01:15.730 new platform called MMP AI Studio. Right, which 00:01:15.730 --> 00:01:18.719 is quite a leap. It really is. So our mission 00:01:18.719 --> 00:01:22.540 today is to explore how this specific AI architecture 00:01:22.540 --> 00:01:26.400 is fundamentally altering the economics and really 00:01:26.400 --> 00:01:29.439 the underlying mechanics of retail video production. 00:01:29.609 --> 00:01:32.829 I mean, we are talking about a massive structural 00:01:32.829 --> 00:01:35.629 shift for both employee training and consumer 00:01:35.629 --> 00:01:37.890 e -commerce. It's a profound shift in how retail 00:01:37.890 --> 00:01:39.650 operates. And what's fascinating here is that 00:01:39.650 --> 00:01:41.269 we really need to understand right out of the 00:01:41.269 --> 00:01:43.290 gate that the scenario you just described. The 00:01:43.290 --> 00:01:45.409 Monday drop with the late video. Yeah, exactly. 00:01:45.489 --> 00:01:47.430 That is not a resource problem. I mean, it's 00:01:47.430 --> 00:01:48.870 not a matter of your learning and development 00:01:48.870 --> 00:01:51.629 team, your L &D folks just, you know, failing 00:01:51.629 --> 00:01:53.489 to work hard enough. Oh, definitely not. They're 00:01:53.489 --> 00:01:55.930 usually overworked as it is. Totally. It is a 00:01:55.930 --> 00:01:58.730 fundamentally broken model colliding with modern 00:01:58.730 --> 00:02:01.680 retail. OK, let's unpack the mechanics of that 00:02:01.680 --> 00:02:04.000 collision. Because if you look at the blog post 00:02:04.000 --> 00:02:07.719 in our sources, they framed the root cause so 00:02:07.719 --> 00:02:11.300 clearly. The timeline for retail has just completely 00:02:11.300 --> 00:02:13.840 changed over the last decade. Oh, beyond recognition. 00:02:14.219 --> 00:02:16.259 Right. Product cycles have accelerated to this 00:02:16.259 --> 00:02:19.259 breakneck pace. We went from having four distinct 00:02:19.259 --> 00:02:24.389 seasons a year to weekly drops, rapid -fire collaborations, 00:02:24.810 --> 00:02:26.629 constant inventory shifts. And the underlying 00:02:26.629 --> 00:02:28.810 friction here is that while the retail calendar 00:02:28.810 --> 00:02:31.469 moved to a hyper -accelerated weekly cadence, 00:02:32.229 --> 00:02:34.710 traditional video production just remains stubbornly 00:02:34.710 --> 00:02:37.349 stuck in the quarterly era. It's so true. I mean, 00:02:37.789 --> 00:02:39.189 traditional production was designed for a time 00:02:39.189 --> 00:02:41.710 when assortments changed maybe four times a year. 00:02:42.129 --> 00:02:45.030 You simply cannot use a quarterly tool to solve 00:02:45.030 --> 00:02:47.530 a weekly problem. No, you can't. And the friction 00:02:47.530 --> 00:02:50.889 becomes so obvious when you break down the traditional 00:02:50.889 --> 00:02:53.909 bottlenecks, the detail in the sources. I mean, 00:02:53.909 --> 00:02:55.870 if you want to make a high -quality video today, 00:02:56.449 --> 00:02:58.909 it is a massive logistical chain. It's a nightmare. 00:02:59.310 --> 00:03:02.129 You write a script. You route it for stakeholder 00:03:02.129 --> 00:03:05.370 approval. You schedule talent, you know, finding 00:03:05.370 --> 00:03:08.270 actors or internal subject matter experts. Then 00:03:08.270 --> 00:03:10.610 you have to book physical studio time. Hire a 00:03:10.610 --> 00:03:12.919 lighting and sound crew. Right. You shoot the 00:03:12.919 --> 00:03:15.680 video, and then you just wait weeks while an 00:03:15.680 --> 00:03:17.819 editor cuts the footage, color corrects, adds 00:03:17.819 --> 00:03:20.319 graphics. And every single link in that chain 00:03:20.319 --> 00:03:24.560 represents time. Right. Time and thousands of 00:03:24.560 --> 00:03:27.560 dollars in capital expenditure per single video. 00:03:27.639 --> 00:03:30.400 It is highly manual. It's a highly serialized 00:03:30.400 --> 00:03:32.280 process. You can't even start step three until 00:03:32.280 --> 00:03:34.659 step two is perfectly finished. It makes me think 00:03:34.659 --> 00:03:39.159 of an analogy to really illustrate the absurdity 00:03:39.159 --> 00:03:41.479 of using that old model today. Let's hear it. 00:03:41.550 --> 00:03:44.169 So relying on traditional video production for 00:03:44.169 --> 00:03:46.870 fast fashion weekly product drops, it's like 00:03:46.870 --> 00:03:49.590 hiring a master oil painter to paint a bespoke 00:03:49.590 --> 00:03:51.449 portrait of your storefront every single week 00:03:51.449 --> 00:03:54.289 just to see who walked in versus just installing 00:03:54.289 --> 00:03:56.210 an automated security camera. I mean, one is 00:03:56.210 --> 00:03:58.909 a slow, expensive piece of art. The other is 00:03:58.909 --> 00:04:01.849 an automated scalable system. built for immediate 00:04:01.849 --> 00:04:04.210 data capture. They just operate on entirely different 00:04:04.210 --> 00:04:06.590 paradigms. Yeah. That is a highly accurate way 00:04:06.590 --> 00:04:09.210 to look at it. You are shifting from bespoke 00:04:09.210 --> 00:04:12.389 creation to algorithmic assembly. And the blog 00:04:12.389 --> 00:04:15.030 post actually identifies the consequence of failing 00:04:15.030 --> 00:04:17.610 to make this shift. They call it the training 00:04:17.610 --> 00:04:20.129 content gap. The training content gap. Right. 00:04:20.350 --> 00:04:23.350 Because when the physical product outpaces the 00:04:23.350 --> 00:04:26.250 informational content, the ripple effects across 00:04:26.250 --> 00:04:29.470 a retail brand are just devastating. Yeah. You 00:04:29.470 --> 00:04:31.819 end up with a gro - back log of products just 00:04:31.819 --> 00:04:35.379 sitting on shelves and an army of associates 00:04:35.379 --> 00:04:38.300 who are expected to sell them but lack any actual 00:04:38.300 --> 00:04:41.490 tools or knowledge to do it well. which naturally 00:04:41.490 --> 00:04:44.029 leads to a massive drop in confidence. Oh, absolutely. 00:04:44.250 --> 00:04:46.889 And from a psychological standpoint, an unconfident 00:04:46.889 --> 00:04:49.509 associate is an immediate liability. Customers 00:04:49.509 --> 00:04:51.750 can smell it, right? They sense hesitation instantly. 00:04:52.250 --> 00:04:54.990 If a shopper asks a really highly specific question 00:04:54.990 --> 00:04:57.750 about a garment's waterproofing and the associate 00:04:57.750 --> 00:05:00.310 guesses or stumbles, the trust is broken, and 00:05:00.310 --> 00:05:02.810 that leads directly to a lost sale. But is the 00:05:02.810 --> 00:05:05.879 alternative just like... letting associates guess 00:05:05.879 --> 00:05:07.699 what the product's features are? Historically, 00:05:07.879 --> 00:05:10.339 yes, or giving them a dense packet of paper they'll 00:05:10.339 --> 00:05:14.339 never read. Ugh, right. And the downstream consequence 00:05:14.339 --> 00:05:16.620 of that gap doesn't just hurt the physical brick 00:05:16.620 --> 00:05:19.360 and mortar store either. The sources point out 00:05:19.360 --> 00:05:22.100 this exact same gap exists on the e -commerce 00:05:22.100 --> 00:05:25.199 side. Oh, for sure. Product pages end up lagging 00:05:25.199 --> 00:05:27.379 behind the actual launches. because they are 00:05:27.379 --> 00:05:30.779 waiting on that exact same slow production cycle 00:05:30.779 --> 00:05:34.500 to create rich, engaging video content. By the 00:05:34.500 --> 00:05:36.939 time the video is live on the website, the optimal 00:05:36.939 --> 00:05:39.319 window to convert those online customers has 00:05:39.319 --> 00:05:41.519 already completely closed. Yeah, you've essentially 00:05:41.519 --> 00:05:44.519 paid to manufacture a product, paid to ship it 00:05:44.519 --> 00:05:46.660 to stores, paid for the marketing to drive foot 00:05:46.660 --> 00:05:49.019 traffic, and then you just stumble at the very 00:05:49.019 --> 00:05:51.360 last inch of the marathon. Just because the human 00:05:51.360 --> 00:05:53.220 being interacting with a customer didn't have 00:05:53.220 --> 00:05:56.759 the context to close the deal. Exactly. content 00:05:56.759 --> 00:05:59.800 gap causing massive operational friction. Which 00:05:59.800 --> 00:06:03.259 brings us to how MMP AI Studio actually functions 00:06:03.259 --> 00:06:05.480 to bypass this entirely. Like how do we get away 00:06:05.480 --> 00:06:07.300 from the blank page problem? Right, the hardest 00:06:07.300 --> 00:06:09.959 part is always starting. Yeah. I was digging 00:06:09.959 --> 00:06:12.519 into the product page from our sources and they 00:06:12.519 --> 00:06:15.959 detail how a user initiates a project. It's basically 00:06:15.959 --> 00:06:19.259 URL magic. The ingestion phase. This is where 00:06:19.259 --> 00:06:22.439 the workful paradigm completely inverts. Right. 00:06:22.500 --> 00:06:24.800 You don't have to write a massive creative brief 00:06:24.800 --> 00:06:28.860 or gather a giant folder of assets. A user literally 00:06:28.860 --> 00:06:32.079 pastes an existing website URL, like a live product 00:06:32.079 --> 00:06:34.980 page, straight into the MMP AI Studio platform. 00:06:35.079 --> 00:06:37.699 Not just copy and paste. Yeah. And the AI then 00:06:37.699 --> 00:06:40.240 automatically scrapes that page. But it's not 00:06:40.240 --> 00:06:42.819 just a copy paste job on the backend. It uses 00:06:42.819 --> 00:06:46.220 natural language processing to extract the product 00:06:46.220 --> 00:06:49.040 names, the descriptions, the key selling features, 00:06:49.439 --> 00:06:52.079 and the high resolution images. It treats your 00:06:52.079 --> 00:06:55.040 existing digital footprint as the raw structured 00:06:55.040 --> 00:06:57.420 data for video production. And from what I understand 00:06:57.420 --> 00:06:59.620 the mechanics, it actually separates the marketing 00:06:59.620 --> 00:07:02.100 fluff from the core specs. Oh, that's crucial. 00:07:02.379 --> 00:07:04.259 Right. It knows the difference between a sentence 00:07:04.259 --> 00:07:07.540 that says evokes the breezy spirit of the Mediterranean. 00:07:07.800 --> 00:07:10.259 in a sentence that says, constructed from a 100 00:07:10.259 --> 00:07:12.720 % moisture wicking merino wool. It grabs the 00:07:12.720 --> 00:07:15.399 wool part. Exactly, it grabs the wool part. Here's 00:07:15.399 --> 00:07:18.600 where it gets really interesting. Then, it auto 00:07:18.600 --> 00:07:21.379 -generates a complete script, maps it out on 00:07:21.379 --> 00:07:24.379 a visual timeline, sequences the entire video, 00:07:24.759 --> 00:07:27.420 creates animated product visuals, and lays down 00:07:27.420 --> 00:07:30.879 a professionally scored... And we really need 00:07:30.879 --> 00:07:33.139 to pause here and look at the operational leverage 00:07:33.139 --> 00:07:36.980 this provides. David Harouche, the CEO of Multimedia 00:07:36.980 --> 00:07:39.000 Plus, is quoted in the press release saying, 00:07:42.859 --> 00:07:45.300 Changing the math. It's a bold claim, but the 00:07:45.300 --> 00:07:47.600 mechanics really do back it up. They do. Because 00:07:47.600 --> 00:07:49.819 you are compressing what traditionally took weeks 00:07:49.819 --> 00:07:51.959 of multi -departmental coordination into minutes 00:07:51.959 --> 00:07:55.060 of backend processing. This completely redefines 00:07:55.060 --> 00:07:57.339 the roles of your personnel. How so? Well, your 00:07:57.339 --> 00:07:59.560 training managers and your L &D leaders are no 00:07:59.560 --> 00:08:01.699 longer acting as stressed -out production assistants 00:08:01.699 --> 00:08:03.699 trying to wrangle actor schedules and lighting 00:08:03.699 --> 00:08:06.120 crews. Oh, thank goodness. Right? The AI handles 00:08:06.120 --> 00:08:08.860 the assembly. The human instantly becomes an 00:08:08.860 --> 00:08:11.459 editor and a director. which frees up their cognitive 00:08:11.459 --> 00:08:13.620 load to actually focus on the strategy of the 00:08:13.620 --> 00:08:15.899 training. They can review what the AI generated, 00:08:16.220 --> 00:08:18.399 tweak the phrasing, ensure the learning objectives 00:08:18.399 --> 00:08:21.019 are met, and just hit publish. It shifts the 00:08:21.019 --> 00:08:23.540 human effort from the tedious logistics of creation 00:08:23.540 --> 00:08:26.199 to the high -value work of curation and strategic 00:08:26.199 --> 00:08:28.720 alignment. Let's talk about the on -camera talent, 00:08:28.899 --> 00:08:31.639 because automated scripts and slick motion graphics 00:08:31.639 --> 00:08:35.039 are great, but retail is inherently human. Very 00:08:35.039 --> 00:08:38.379 human. So if you replace human actors with AI, 00:08:38.960 --> 00:08:42.899 How do you avoid the video feeling entirely robotic, 00:08:43.100 --> 00:08:45.340 like you were just watching a spreadsheet read 00:08:45.340 --> 00:08:47.980 itself out loud? That has been the primary barrier 00:08:47.980 --> 00:08:50.639 to entry for early AI video, you know, the uncanny 00:08:50.639 --> 00:08:52.379 valley effect. Yeah, where it just feels a bit 00:08:52.379 --> 00:08:56.299 off. Exactly. It felt sterile, generic, completely 00:08:56.299 --> 00:08:59.019 disconnected from the brand's reality. But the 00:08:59.019 --> 00:09:01.700 sources outline how MMP addresses this through 00:09:01.700 --> 00:09:03.820 what they call the virtual presenter library. 00:09:04.080 --> 00:09:06.649 This part is fascinating to me. You generate 00:09:06.649 --> 00:09:09.809 a roster of AI on -camera hosts, and each one 00:09:09.809 --> 00:09:12.610 has a unique look, personality, and voice. But 00:09:12.610 --> 00:09:15.149 the mechanism that actually grounds this, and 00:09:15.149 --> 00:09:17.409 this is so specific to the psychology of retail, 00:09:18.269 --> 00:09:21.429 is the environment rendering. These AI -generated 00:09:21.429 --> 00:09:23.990 hosts don't just stand in front of a generic 00:09:23.990 --> 00:09:26.289 white background or a fake news desk. Right, 00:09:26.350 --> 00:09:29.320 which looks so cheap. Totally. This system actually 00:09:29.320 --> 00:09:32.259 renders them standing inside your brand's actual 00:09:32.259 --> 00:09:35.220 physical store environments, wearing your brand's 00:09:35.220 --> 00:09:37.559 specific uniform guidelines. That is a vital 00:09:37.559 --> 00:09:40.460 detail for cognitive immersion. If an associate 00:09:40.460 --> 00:09:42.419 is standing in a back room watching a training 00:09:42.419 --> 00:09:45.360 video, the context needs to mirror their reality. 00:09:45.690 --> 00:09:48.029 Yeah, if the avatar is wearing the same lanyard, 00:09:48.210 --> 00:09:50.629 the same apron, standing in front of the exact 00:09:50.629 --> 00:09:52.889 same fixture, the associate is about to walk 00:09:52.889 --> 00:09:55.110 out and clean. The cognitive distance between 00:09:55.110 --> 00:09:57.870 the training and the application drops to zero. 00:09:57.950 --> 00:10:00.970 It maintains absolute brand consistency. It makes 00:10:00.970 --> 00:10:03.370 the content feel native to the company culture 00:10:03.370 --> 00:10:06.289 rather than something outsourced to a third party 00:10:06.289 --> 00:10:08.549 agency that just doesn't understand the vibe 00:10:08.549 --> 00:10:10.929 of the brand. Precisely. But if we want to talk 00:10:10.929 --> 00:10:13.909 about true global scale, we have to examine the 00:10:13.909 --> 00:10:16.730 multi -language video generation. Because for 00:10:16.730 --> 00:10:19.149 any retail executive managing an international 00:10:19.149 --> 00:10:22.190 footprint, localization is the exact point where 00:10:22.190 --> 00:10:25.009 the entire production schedule usually just collapses. 00:10:25.269 --> 00:10:27.149 Oh, absolutely. The traditional localization 00:10:27.149 --> 00:10:29.669 process is a nightmare. You have to translate 00:10:29.669 --> 00:10:32.370 the script, hire entirely new regional voice 00:10:32.370 --> 00:10:34.870 actors, bring them into a booth, and then have 00:10:34.870 --> 00:10:37.659 an editor manually adjust the original video 00:10:37.659 --> 00:10:40.559 to try and match the new audio track to the old 00:10:40.559 --> 00:10:42.740 visuals. It's why so many brands simply don't 00:10:42.740 --> 00:10:44.679 localize. They just can't afford the time or 00:10:44.679 --> 00:10:47.100 the money. Right. But with this platform, the 00:10:47.100 --> 00:10:50.580 sources describe a one -click localization architecture. 00:10:50.620 --> 00:10:53.340 Just one click. One click. You generate the core 00:10:53.340 --> 00:10:56.460 video in English, then you select a target language, 00:10:56.679 --> 00:10:59.399 and the AI translates the dialogue natively, 00:10:59.700 --> 00:11:02.340 completely re -voices the audio using the avatar's 00:11:02.340 --> 00:11:04.700 voice model. Wait, really? It keeps the same 00:11:04.700 --> 00:11:07.460 voice. Yes, and it dynamically re -renders the 00:11:07.460 --> 00:11:09.580 visual presenter's mouth movements to match the 00:11:09.580 --> 00:11:12.620 new language and localizes all the on -screen 00:11:12.620 --> 00:11:14.860 graphics and captions? In a matter of minutes. 00:11:15.320 --> 00:11:18.220 Minutes, with zero incremental production cost. 00:11:18.440 --> 00:11:21.019 Think about the strategic leverage this hands 00:11:21.019 --> 00:11:24.480 a global brand. Historically, because of strict 00:11:24.480 --> 00:11:26.799 profit and loss constraints, you know, the P 00:11:26.799 --> 00:11:29.919 &L, executives had to make painful budgetary 00:11:29.919 --> 00:11:33.440 choices. Like, do we spend $20 ,000 to produce 00:11:33.440 --> 00:11:35.980 this vital training video in English, or do we 00:11:35.980 --> 00:11:39.039 do it in Spanish? Exactly. Or worse, you just 00:11:39.039 --> 00:11:42.259 produce it in English, slap some tiny hard -to 00:11:42.259 --> 00:11:44.399 -read localized subtitles on it, and just hope 00:11:44.399 --> 00:11:46.740 your teams in Mexico or Germany can parse it 00:11:46.740 --> 00:11:49.200 out. Which is so common. But when the marginal 00:11:49.200 --> 00:11:51.779 cost of localization drops to effectively zero, 00:11:52.179 --> 00:11:54.700 you don't choose anymore. You instantly produce 00:11:54.700 --> 00:11:58.029 English, Spanish, French, German and Japanese. 00:11:58.309 --> 00:12:00.450 The bottleneck of language is entirely removed 00:12:00.450 --> 00:12:02.769 from the PNL equation. This actually allows brands 00:12:02.769 --> 00:12:05.309 to test marketing and training in secondary markets, 00:12:05.549 --> 00:12:07.950 say expanding into Southeast Asia or Eastern 00:12:07.950 --> 00:12:10.370 Europe, markets that previously couldn't justify 00:12:10.370 --> 00:12:13.450 the upfront ROI of localized video production. 00:12:13.649 --> 00:12:15.169 OK, I have to play the skeptic here for a minute. 00:12:15.210 --> 00:12:17.590 Go for it. Because if I'm a retail executive 00:12:17.590 --> 00:12:19.429 listening to this, I'm hearing all this about 00:12:19.429 --> 00:12:22.669 AI video generation and I'm thinking, look, there 00:12:22.669 --> 00:12:26.379 are dozens of generic, free or very cheap AI 00:12:26.379 --> 00:12:28.419 video generators out on the internet right now. 00:12:28.700 --> 00:12:31.500 Oh, tons of them. So why would I pay for an enterprise 00:12:31.500 --> 00:12:34.419 purpose -built platform? Couldn't a scrappy marketing 00:12:34.419 --> 00:12:36.700 intern just use one of those random off -the 00:12:36.700 --> 00:12:39.659 -shelf AI tools to make these videos? It is the 00:12:39.659 --> 00:12:42.960 most logical executive pushback. Why buy the 00:12:42.960 --> 00:12:46.279 specialized tool when a generic large language 00:12:46.279 --> 00:12:48.940 model, an LLM, is sitting right there? Right. 00:12:49.100 --> 00:12:51.120 But if we connect this to the bigger picture 00:12:51.120 --> 00:12:54.139 of the operational realities of retail execution, 00:12:54.679 --> 00:12:57.399 relying on generic AI is incredibly risky. I 00:12:57.399 --> 00:13:00.200 hope so. The blog post explicitly breaks down 00:13:00.200 --> 00:13:02.799 why generic tools fail in this specific context. 00:13:03.360 --> 00:13:05.820 Retail microlearning operates under very strict 00:13:05.820 --> 00:13:08.580 non -negotiable requirements. The primary one 00:13:08.580 --> 00:13:11.539 is absolute product accuracy. Ah, because generic 00:13:11.539 --> 00:13:13.940 AI is infamous for hallucinating, right? Like 00:13:13.940 --> 00:13:16.179 confidently making up facts that look structurally 00:13:16.179 --> 00:13:18.789 plausible but are completely wrong. Exactly. 00:13:19.230 --> 00:13:21.830 Now imagine your intern uses a generic AI to 00:13:21.830 --> 00:13:24.590 generate a training video. The AI confidently 00:13:24.590 --> 00:13:26.789 scripts and presents a video telling your frontline 00:13:26.789 --> 00:13:29.710 associates that a new sweater collection is 100 00:13:29.710 --> 00:13:32.690 % cashmere when the actual product is a synthetic 00:13:32.690 --> 00:13:35.490 wool blend. Oh wow, that's a disaster. But wait, 00:13:35.669 --> 00:13:38.110 LLMs are getting better every day. Couldn't that 00:13:38.110 --> 00:13:40.509 intern just use a really strict negative prompt? 00:13:40.830 --> 00:13:43.549 Something like, under no circumstances should 00:13:43.549 --> 00:13:46.360 you hallucinate facts about this product. Prompt 00:13:46.360 --> 00:13:48.899 engineering is not a systemic safeguard. It is 00:13:48.899 --> 00:13:52.059 a polite request to a probabilistic engine. A 00:13:52.059 --> 00:13:54.789 polite request. I like that. Yeah. Generic AI 00:13:54.789 --> 00:13:57.009 pulls from the entire open internet to guess 00:13:57.009 --> 00:13:59.590 the most likely next word. It doesn't inherently 00:13:59.590 --> 00:14:02.909 understand your product. MMP AI Studio, however, 00:14:02.990 --> 00:14:05.370 uses a closed -loop data architecture. Meaning 00:14:05.370 --> 00:14:07.370 it restricts the universe of information the 00:14:07.370 --> 00:14:09.730 AI is allowed to pull from. Precisely. It is 00:14:09.730 --> 00:14:12.309 constrained via strict API guardrails to only 00:14:12.309 --> 00:14:14.570 utilize the specific product data it scraped 00:14:14.570 --> 00:14:17.129 from your URL or that you explicitly uploaded. 00:14:17.509 --> 00:14:20.629 There is zero room for error when you are teaching 00:14:20.629 --> 00:14:22.929 an associate what to tell a paying customer. 00:14:22.889 --> 00:14:26.769 It is built for absolute factual fidelity, not 00:14:26.769 --> 00:14:28.590 just rendering a pretty picture. Which makes 00:14:28.590 --> 00:14:31.210 total sense because the liability of a hallucination 00:14:31.210 --> 00:14:34.350 in retail is massive. If the associate tells 00:14:34.350 --> 00:14:36.870 the customer it's cashmere, the customer buys 00:14:36.870 --> 00:14:39.710 it for a premium, takes it home, realizes it's 00:14:39.710 --> 00:14:43.110 a blend, feels completely lied to, returns it, 00:14:43.289 --> 00:14:46.690 and leaves a scathing review online. Yep. One 00:14:46.690 --> 00:14:50.110 AI hallucination just permanently damaged a customer 00:14:50.110 --> 00:14:52.690 lifetime relationship. Unbelievable. Beyond the 00:14:52.690 --> 00:14:54.730 accuracy of the data, there is the complexity 00:14:54.730 --> 00:14:57.789 of retail training scenarios. The sources highlight 00:14:57.789 --> 00:15:00.690 a specific capability, role -play and scenario 00:15:00.690 --> 00:15:03.029 videos. Yes, I wanted to dig into this because 00:15:03.029 --> 00:15:05.169 the platform allows you to generate scenes using 00:15:05.169 --> 00:15:07.940 two distinct. AI actors interacting with each 00:15:07.940 --> 00:15:10.139 other on screen. Which is incredibly difficult 00:15:10.139 --> 00:15:12.820 to do with generic off -the -shelf AI. I mean, 00:15:12.820 --> 00:15:14.620 generating one person talking to a camera is 00:15:14.620 --> 00:15:17.460 relatively easy. But generating a multi -turn 00:15:17.460 --> 00:15:20.120 back and forth conversation between two avatars 00:15:20.120 --> 00:15:23.620 that visually models a complex customer interaction, 00:15:24.159 --> 00:15:26.720 like handling a return dispute or walking through 00:15:26.720 --> 00:15:29.279 an upselling technique, that requires a level 00:15:29.279 --> 00:15:31.799 of sequencing that generic tools simply aren't 00:15:31.799 --> 00:15:35.100 built for. Because you can't teach the soft skills 00:15:35.100 --> 00:15:38.610 of retail. like tone, pacing, conflict resolution 00:15:38.610 --> 00:15:41.289 with a bulleted list on a break room wall, you 00:15:41.289 --> 00:15:43.759 have to visually model the behavior. And even 00:15:43.759 --> 00:15:46.000 with all this automation, the control architecture 00:15:46.000 --> 00:15:49.440 remains firmly with the human experts. The sources 00:15:49.440 --> 00:15:52.100 note that training teams retain full editorial 00:15:52.100 --> 00:15:54.820 control before anything goes live. You aren't 00:15:54.820 --> 00:15:56.799 just crossing your fingers and hoping the AI 00:15:56.799 --> 00:15:59.299 spits out something usable. You can edit the 00:15:59.299 --> 00:16:00.799 script right there in the browser interface. 00:16:01.039 --> 00:16:03.779 Or export it to a Word document to route it through 00:16:03.779 --> 00:16:06.399 your legal or compliance departments for review 00:16:06.399 --> 00:16:09.039 before the final video even renders. Exactly. 00:16:09.100 --> 00:16:11.379 You are firmly in the driver's seat. And the 00:16:11.379 --> 00:16:13.250 workflow efficiency continues. right through 00:16:13.250 --> 00:16:16.610 to the distribution phase, which is huge. Once 00:16:16.610 --> 00:16:19.649 the video is perfect, MMP publishes it directly 00:16:19.649 --> 00:16:22.049 into their flagship mobile -first platform called 00:16:22.049 --> 00:16:24.710 INCITE. Which is a crucial step. It really is. 00:16:25.009 --> 00:16:27.809 It bypasses the messy reality of downloading 00:16:27.809 --> 00:16:31.090 massive MP4 files, uploading them to a clunky 00:16:31.090 --> 00:16:33.350 corporate internet, and then just hoping someone 00:16:33.350 --> 00:16:35.909 clicks the link. Which they never do. Right. 00:16:36.230 --> 00:16:38.210 It pushes the video directly to the frontline 00:16:38.210 --> 00:16:40.789 associate's mobile devices. where they already 00:16:40.789 --> 00:16:43.350 work and communicate. It solves the last mall 00:16:43.350 --> 00:16:46.470 distribution problem just as elegantly as it 00:16:46.470 --> 00:16:48.570 solves the creation problem. But we would be 00:16:48.570 --> 00:16:50.850 doing the listeners a disservice if we limited 00:16:50.850 --> 00:16:53.230 our view of this technology strictly to internal 00:16:53.230 --> 00:16:55.850 training. The economic leverage here extends 00:16:55.850 --> 00:16:58.210 far beyond the break room. Which brings us to 00:16:58.210 --> 00:17:01.049 the ultimate two birds, one stone scenario detailed 00:17:01.049 --> 00:17:03.809 in the sources, the crossover into e -commerce. 00:17:04.029 --> 00:17:06.809 The underlying logic is elegant. It is the exact 00:17:06.809 --> 00:17:10.130 same core data. Right. The same validated product 00:17:10.130 --> 00:17:12.769 data, the features, the benefits, the high -res 00:17:12.769 --> 00:17:15.069 visuals that you use to train your internal associate 00:17:15.069 --> 00:17:17.720 can be instantly repurposed. The platform can 00:17:17.720 --> 00:17:20.579 take that exact same data set and render shoppable, 00:17:20.880 --> 00:17:22.900 consumer -facing videos designed specifically 00:17:22.900 --> 00:17:25.519 for your external e -commerce product pages, 00:17:25.779 --> 00:17:27.700 your digital ad campaigns, your social media 00:17:27.700 --> 00:17:31.289 channels. It creates a massive ROI multiplier. 00:17:31.769 --> 00:17:34.390 You are leveraging a single core effort to drive 00:17:34.390 --> 00:17:37.730 both employee competency and direct consumer 00:17:37.730 --> 00:17:40.349 conversion. That's incredibly efficient. It is. 00:17:40.509 --> 00:17:42.630 Jodi Harouche, the president and chief creative 00:17:42.630 --> 00:17:45.349 officer of MMP, is quoted in the press release 00:17:45.349 --> 00:17:48.269 capturing this dynamic perfectly. She notes that 00:17:48.269 --> 00:17:51.410 this platform puts broadcast quality video production 00:17:51.410 --> 00:17:53.430 directly in the hands of the people who know 00:17:53.430 --> 00:17:56.009 the products best. The people who actually understand 00:17:56.009 --> 00:17:58.529 the merchandising strategy rather than a third 00:17:58.529 --> 00:18:00.930 -party production house that just knows how to 00:18:00.930 --> 00:18:03.650 set up lighting, it removes the middleman completely. 00:18:04.250 --> 00:18:06.950 And ultimately for a retail executive analyzing 00:18:06.950 --> 00:18:09.809 their P &L, all of this technological capability 00:18:09.809 --> 00:18:12.529 ladders up to a single critical performance metric 00:18:12.529 --> 00:18:15.269 mentioned in the blog post, speed to competency. 00:18:15.430 --> 00:18:17.819 Speed to competency, let's define that. How rapidly 00:18:17.819 --> 00:18:19.779 can you take a frontline worker from completely 00:18:19.779 --> 00:18:22.339 ignorant about a new product to confidently selling 00:18:22.339 --> 00:18:24.980 it on the floor? Exactly. Faster content generation 00:18:24.980 --> 00:18:27.140 means you can actually keep up with weekly product 00:18:27.140 --> 00:18:30.019 drops. Relevant, visually engaging video training 00:18:30.019 --> 00:18:32.460 builds that competency infinitely faster than 00:18:32.460 --> 00:18:35.920 handing an associate a dense PDF manual. Definitely 00:18:35.920 --> 00:18:38.619 and that speed directly translates to associates 00:18:38.619 --> 00:18:41.440 who are confident on day one and confident associates 00:18:41.440 --> 00:18:44.099 directly measurably increase average transaction 00:18:44.099 --> 00:18:47.059 value and elevate the in -store customer experience 00:18:47.059 --> 00:18:50.480 So what does this all mean if we step back and 00:18:50.480 --> 00:18:52.559 look at the entire landscape? We've mapped out 00:18:52.559 --> 00:18:57.039 today MMP AI Studio is taking retail video production 00:18:57.039 --> 00:19:00.900 from a highly serialized, costly, weeks -long 00:19:00.900 --> 00:19:04.339 studio process and transforming it into an instant, 00:19:04.480 --> 00:19:07.319 scalable, closed -loop workflow. A total paradigm 00:19:07.319 --> 00:19:09.940 shift. By utilizing existing digital footprints, 00:19:10.259 --> 00:19:13.400 it's generating customized, multilingual, factually 00:19:13.400 --> 00:19:16.240 accurate content. And it's doing this simultaneously 00:19:16.240 --> 00:19:19.119 for the backroom where your team trains and the 00:19:19.119 --> 00:19:21.220 online storefront where your customers buy. It 00:19:21.220 --> 00:19:22.920 is taking the friction completely out of the 00:19:22.920 --> 00:19:25.140 system. It is entirely rewriting the operational 00:19:25.140 --> 00:19:27.500 playbook. But it also raises a deeper, perhaps 00:19:27.500 --> 00:19:29.079 more challenging question. This is something 00:19:29.079 --> 00:19:31.819 I think every retail executive listening needs 00:19:31.819 --> 00:19:34.339 to really ponder as this type of instantaneous 00:19:34.339 --> 00:19:37.339 AI capability becomes the baseline standard across 00:19:37.339 --> 00:19:39.359 the industry. Oh, I love a good strategic thought 00:19:39.359 --> 00:19:41.279 experiment. Where does this lead us? Let's follow 00:19:41.279 --> 00:19:45.660 the trajectory. If AI completely solves the speed 00:19:45.660 --> 00:19:48.880 to competency bottleneck for everyone, if every 00:19:48.880 --> 00:19:52.140 single retail associate worldwide, across all 00:19:52.140 --> 00:19:55.160 your competitors, is suddenly perfectly trained 00:19:55.160 --> 00:19:58.279 on every granular product feature the exact second 00:19:58.279 --> 00:20:02.349 a new collection drops, How will your brand differentiate 00:20:02.349 --> 00:20:05.430 its in -store customer experience? Wow. Right. 00:20:05.450 --> 00:20:07.670 When instantaneous, perfect product knowledge 00:20:07.670 --> 00:20:10.569 is commoditized and no longer a unique operational 00:20:10.569 --> 00:20:13.609 advantage, what becomes the new premium in retail? 00:20:13.789 --> 00:20:16.069 That is a phenomenal point. When the knowledge 00:20:16.069 --> 00:20:18.710 itself is ubiquitous, the competitive battlefield 00:20:18.710 --> 00:20:21.890 shifts entirely to human empathy, connection, 00:20:22.130 --> 00:20:23.710 and what you actually do with that knowledge 00:20:23.710 --> 00:20:25.650 in front of the customer. Exactly. You've automated 00:20:25.650 --> 00:20:27.309 the carriage building. Now you have to figure 00:20:27.309 --> 00:20:29.960 out entirely new uncharted places to drive. Think 00:20:29.960 --> 00:20:32.420 on that. Thanks for joining us on this deep dive. 00:20:32.539 --> 00:20:33.319 We'll catch you next time.