The questions a skeptical marketer actually asks

That setup is genuinely good, and if you produce a handful of assets a month with a senior person driving, keep it. Here is the line it cannot cross: stored context makes regeneration cheap; it does not make it reproduction. Every time the session regenerates, the model re-derives the deck from context: similar, never the same. Layout, spacing, type, and structure drift run to run, because generating is the only move the machine has. A trained template here is not a prompt. The approved design's geometry and type scale are stored as data, the model only fills the slots, and "only what you asked to change has changed" is a claim you can literally diff. Regeneration is a talented freelancer redrawing your deck from memory; reproduction is a production line re-running it.

Second, in the session your brand rules are instructions the model usually follows, and the reviewer is you: the model grades its own homework, so every "free" regeneration still costs senior minutes of inspection, and the drift you miss is the drift that ships. Here the checks are code that measures the output (your exact hex, your real fonts, the logo file on record or no logo at all, every claim resolving to a source) and refuses to file a failure. And each regeneration in a session is a new artifact: no version history across runs, no per-block revert, no approval chain, no login-free review link for the person who signs off, no record of what shipped where. The moment sign-off matters, a chat is governance by DM.

Last, the setup you describe is deck-shaped, and the production job is not: carousels and ad sets at exact pixel specs, print-ready sell sheets, email-safe HTML, video cutdowns, published to LinkedIn behind a human approval gate, all under the same guarantees. The honest end state is not either-or. This platform ships an MCP server, so you can point the cowork session you already like at it: the session writes the brief, the production line executes it with everything above enforced. We are not the alternative to how you work. We are what makes it safe to scale past you.

A wrapper asks a model to behave. This platform makes the behavior unavoidable: brand rules, layout fidelity, logo handling, lineage, and the no-AI-tells pass are enforcement gates in the production path, and an asset cannot reach you without passing through them (the ten guarantees on the home page are that list). The other half a wrapper cannot do: the training engine, which decomposes a design your brand already approved and turns it into a runnable template that reproduces at that design's exact geometry. The model underneath is swappable and always will be. The blocks everything is built from, the trained templates, and the enforcement are the product.

It erases products that compete with models on creative quality. We deliberately do not: strategy and creative judgment stay with you, and raw generation comes from the frontier models we run (Claude and Gemini on Vertex AI, swappable). What we own is the layer the models do not: your accumulated library, your trained templates, brand enforcement in code, review and approval, and publishing. Every model improvement makes our output better at no cost to you, which means we appreciate with the models rather than racing them.

Canva is a design tool: excellent at helping a person edit an asset. We are not trying to be a better editor. The job here is different: hold everything your team has ever shipped in one searchable record, learn the designs your brand already approved, produce new work locked to them, gate it through review, and publish it, with a team that may never open a design tool at all. Where Canva and tools like it keep improving single-asset editing, that helps our customers too; the account-level layer (your library, your templates, your enforcement, your approvals) is the part nobody else is building for you.

The slop problem is exactly why this product refuses to sell volume. Everything produced here is built from your own material (your white papers, your webinars, your case studies) in your own approved designs, and a mechanical pass strips the AI tells (placeholder text, giveaway punctuation, broken generated images) before anything is filed. The test we hold ourselves to: two companies running the identical request from comparable sources must get visibly different work, each recognizably theirs. If output ever reads like everyone else's feed, we have failed at the one thing we exist for.

The menu on the pricing page is the honest contract: sales decks, sell sheets and one-pagers, white paper and ebook design, infographics, LinkedIn carousels, social graphic sets, ad sets with variants, marketing emails, landing pages, video cutdowns from your footage, and templates trained from your approved designs. The menu grows only when the platform genuinely produces the item; a menu line the engine cannot honor would be a lie on a pricing page, so we do not write one.

Depends what "original" means, so here is the exact line. New concepts inside your brand (a campaign angle, a fresh layout within your design system, the next twenty ad variants): yes, that is the job. A new visual identity (a logo, a net-new brand, an original illustration style): no, permanently. That is human creative work, it is the part you should own, and a tool that invents it for you is how every company ends up looking the same. We execute on yours.

Less than you expect, and never zero. The copy is built from your own filed material with lineage recorded as it is written, and a reference that does not resolve blocks the asset from being filed, so the classic AI failure (confident invented claims) is structurally blocked rather than hoped away. You should still expect an editing pass; that is why everything arrives as live, editable text with version history rather than a flat picture, and why the top two rungs put a human review before work reaches you at all.

No. An agency sells judgment and production bundled; keep the judgment wherever you trust it, including your agency. What this replaces is the production half of the invoice: the hundredth version, the resize, the reformat, the variant set. Plenty of teams will run both, with the retainer shrinking toward strategy. What we do not do is sell to agencies; the customer here is the internal team.

Yes, on one tier specifically: Managed Production. You send briefs and finished work comes back; nobody on your team has to learn anything. Onboarding builds your brand kit and first templates from whatever you have, even if that is just a website and a pitch deck. What it does not include is a marketing strategy; when you want that thought through, it is consulting, priced separately and honestly.

Four gates, in order. A mechanical check validates what is simply true or false about a design (real fonts, exact hex, logo file on record or no logo at all). A brand check scores color, type, logo, and copy against your rules, with findings that name the value measured. Lineage is recorded per claim as the asset is written, and a reference that does not resolve is never filed. And nothing publishes without a person: every publish path sits behind human approval, with review links your approver can open without an account. The honest framing: the gates make failures rare and catchable, and the human approval is there because "rare" is not "never".

No. Model calls run directly on Google Vertex AI enterprise endpoints, which by policy do not train on customer content, with no third-party routing in between. Your data lives in PostgreSQL with row-level security enforcing isolation between organizations on every query, verified by an automated check in our build pipeline. Identity is Auth0. The full posture, including our subprocessors and their certifications, is published at /trust. What the platform learns about you (styles, rules, templates, examples) is visible in the product, editable, and deletable, never a black box.

Not yet, and we will not imply otherwise. Every subprocessor we run on holds its own certifications (SOC 2 Type II and beyond; the table is at /trust), we maintain Technology E&O with Cyber liability coverage, and a DPA is available on request. Our own SOC 2 Type I and independent penetration testing are on the compliance roadmap and publish at /trust as each milestone is scheduled. If your security review needs something specific, ask; you will get a straight answer.

Never. Automations produce drafts that land in review; a person ships every single thing that leaves the platform, and there is no setting that removes the human. This is enforced as a code path, not a policy document.

DesignTech AI was designed and built end to end by its founder: a designer, marketer, and agency owner with more than twenty years in B2B, whose agency delivered 2,626 client projects and 12,723 billed hours of exactly this work. That is where the best practices in the platform come from: not a scraped training set, but one practitioner's production discipline written into the prompts, the templates, and the enforcement gates. It is also why, at the founding stage, the person on your calls is the person who built the product.

On vendor risk, plainly: the company is small and early, and the protections are structural rather than promises. Your workspace exports in full at any time, your data lives on audited managed infrastructure (/trust), Technology E&O and Cyber insurance are in force, and terms are monthly. You are never more than one export away from owning everything you made.

You own it all, contractually: everything ingested, every template trained from your designs, and every asset produced in your workspace, including work our team produced for you. Terms are monthly with cancellation effective at the end of the billing month, your library exports in standard formats (PDF, PNG, PPTX, MP4, and more), and everything the platform learned about you can be reviewed and deleted. We would rather earn the renewal than lock the door.

Three tiers, priced on who does the labor: Managed Production at $3,000/month (we deliver finished work through a production queue), Self-Serve at $750/month (your team runs it, brand-locked; the optional QA & Coaching Pack at +$750/month adds a human review gate, monthly coaching, and double the seats and credits), and Enterprise (SSO, custom volume, priced by conversation). AI usage is metered in credits with the pool included in every tier and visible in-app, so the bill does not spike behind your back; overflow is a listed rate, not a surprise. The whole model, including the annual Charter Program rates, is public at /pricing.

Two ways. The free tier runs on your real content with no account required. And the first conversation is an audit, not a pitch: bring one approved design and a list of what you produced last quarter, and we show you, item by item, what the platform absorbs. If the answer is "not much", you will hear that too.

We will not inflate this: the customer base today is early, and that is exactly why the Charter Program exists, with charter pricing locked for as long as you stay in exchange for being early. What we can show you beats a logo wall anyway: the platform producing from your own materials, live on a call, on work you can judge yourself.