teradata. AI Sovereignty · Talk Track Guide

Wave 1 · Draft v0.1 🇬🇧 EN Landing page Executor Open the base Talk Track ↗
Sovereignty Talk Track · Field Guide

A companion, not a replacement.

The Kill Switch parable ships with a rehearsed, beat-by-beat presenter script bundled in this pack. This guide is the Wave 1 sales-play layer on top of it — the outreach choreography, persona-to-beat pairing, the turn to the AI Sovereignty Model, the only moment a product is named, and the workshop hand-off that the base talk track doesn't cover.

↗

Open the base Talk Track first

Every reference in this guide points to a beat of the bundled storyline talk track (anchors #b0–#b7). Have it open in a second tab. The sections below map when to open which Executor persona, what to do between beats, and when to hand off to Strategy or ROI. It does not repeat the base script.

🎙️ Base Talk Track EN 🛡️ AI Sovereignty Model 📖 The Kill Switch
👤Audience CIO · CISO · DPO · GC · Board ⏱Base script ~13 min (bundled) 📄Companion This guide 🌐Languages Parable EN + DE · script EN 🛡️Guardrail No political actors named
Section 01

Before the meeting · Persona-tuned outreach

Goal: Use the Executor's persona essay to identify which of the six places the prospect is publicly exposed on, then send a parable-anchored outreach. The Executor is not a "sales assist" — it is the discovery layer that determines where in the base talk track to dwell.

✉Sequence — Executor first, meeting second

  • 01Open the AI Sovereignty Executor (executor.html). Paste prospect signals. Run Stage 1 to produce the persona essay.
  • 02Read the essay's "public gap" line — it tells you which beat of the base talk track to dwell on. Usually the count (Beat 3) for CIOs; the party further back (Beat 4) for CISOs; residency-vs-sovereignty (Beat 1) for DPOs.
  • 03Run Stage 2 to produce the outreach. Send it with the Kill Switch parable link — the parable is the only asset that travels; the customer will read it and arrive primed.
  • 04The meeting then opens with the customer already carrying the distinction — residency answers where, sovereignty answers who can be told what to do. That changes how hard each beat has to work.

🎯Persona → base talk-track beat pairing

CIO
Dwell on the count. Beat 3 (How many people have to say yes) is their beat: continuity arithmetic on systems they own. Give the count exercise the full silence it needs, and let Beat 5's register land as the operational fix.
CISO
Dwell on the party further back. Beat 4 (The one that decides sits further back) maps to their tier-two supply-chain blind spot. Pair it with the DORA subcontracting-chain requirement when the register comes up.
DPO / GC
Dwell on the two words. Beat 1 (Residency is not sovereignty) is the legal distinction they are paid to hold, and Beat 2's mechanism — an order that reaches through a local contract — is their daily jurisdiction question made concrete.

▶Cue — what to test before the meeting

1Run the parable once on the machine and screen you will present from. The Beat 2 cascade — seven services going dark against the gauge — is the emotional centre of the story; if it stutters, the story stutters.
2Have three tabs open in order: why-the-kill-switch.html (live), why-the-kill-switch-talktrack.html (your script), dist/AI_Sovereignty_Model.html (for the turn — and not before).
3Have this guide open too — the persona pairing above tells you where to dwell. Do not read it aloud.

?Ask the Executor essay first

  1. Which of the six places does the essay identify as this persona's publicly active concern? — That is where you dwell in the base talk track.
  2. Which place is their public gap — the one their published posture never mentions? — That is where the register will surprise them, and where the 2-hour workshop earns its keep.
  3. Does the essay surface a public phrase — a keynote line, a regulatory filing, an interview quote — you can quote back verbatim? If yes, open the meeting with it before touching the parable.

🔀Cross-play routing signals

SIGNAL AThe essay leads with journey and roadmap language — "scaling AI", maturity, operating model. → Route to AI Strategy. Send The Ascent parable as follow-up.
SIGNAL BThe essay leads with AI spend and unit economics — capex on the P&L, consumption pricing, egress. → Route to AI ROI. Send The Cost of More parable as follow-up.
SIGNAL CThe essay says "we already run a sovereign cloud / residency programme". → Do not compete with it. Use their programme as the anchor and position the register as its missing evidence layer — named parties, rehearsed exits.
Transition to Section 2 Now the meeting starts. The base talk track is your storyline. This guide's Section 2 tells you what to do between the beats.
Section 02

During the meeting · Working the base talk track

Goal: Run the base talk track naturally — but use these Play-specific pauses, questions, and interaction cues to make the story land as a conversation about their estate rather than a lecture about a stranger's bad week.

◉Beat-by-beat pairing with the base talk track

Base TT · Beats 0–1 ~2:30
The hero · Residency is not sovereignty
Play addition: After the two columns, pause and ask: "If I asked your board today 'how sovereign is our AI?' — what would they answer?" Do not correct whatever comes back. It is your anchor for the rest of the meeting.
Open base TT · Beat 0
Base TT · Beat 2 ~2:00
The cards went first
Play addition: The cascade runs on its own — seven services go dark against the gauge. Stop speaking and let the room watch. If anyone starts arguing about the case itself, you have over-told it: drop the story, return to the mechanism.
Open base TT · Beat 2
Base TT · Beats 3–4 ~3:30
The count · the one that decides sits further back
Play addition: After the chain, stop and ask the room to do the count for one production AI call. Wait through the silence. Running out at three is the designed outcome, not a failure of their diligence — say so.
Open base TT · Beat 3
Base TT · Beats 5–6 ~3:45
The register · the three moves
Play addition: Walk the counting ladder and ask where they stop — most say three, and saying it out loud is the point. Read the hinge line flat: "The page removed nothing. Not one yes." Then read the three moves off the page — reroute, retain, rehearse. Do not paraphrase them into advice.
Open base TT · Beat 5
Base TT · Beat 7 ~1:30
The turn
Play addition: Slow right down. "Nobody writes that page for an enterprise" is the hand-off sentence — this guide's Section 3 begins the moment it lands. Do not open the model before it has.
Open base TT · Beat 7

▶Live interactivity cues

1In Beat 3, use the chain picker: open Her credit cards, read the chain in order, then click Her bank account. Different thing entirely — same party at the bottom. Say nothing. Let them see it.
2In Beat 6, point at the before/after — REACHES 3 OF 3 → REACHES 1 OF 3. Land "not zero." Then the note under it: nothing given up.
3Do not open the AI Sovereignty Model during the story. The parable never names it before the turn, and neither do you.

?Ask between base TT beats

  1. After Beat 2: "Has a supplier ever told you they were legally prohibited from telling you something? Would you expect to be told if they were?"
  2. After Beat 4: "Does your third-party register go past the parties you contract with directly? What sits behind them?"
  3. After Beat 6: "What do you actually hold — data in a usable export, model weights, keys — versus what you hold a permission to use?"

⚠Objections the base script doesn't cover

"We have a third-party risk register already."
Most do, and they are good at tier one. The question is what sits behind tier two, and whether the register records the parties you never chose. DORA's subcontracting-chain requirement exists precisely because that is where registers stop.
"An enterprise has legal recourse. She did not."
Recourse takes months. The question the story asks is what happens in the meantime, and whether you would even know which systems were affected on day one. Recourse is a remedy; it is not a control.
"We cannot possibly map all of this."
You do not have to map all of it. The number is finite, and one answer usually clears several questions at once, because the same few parties sit behind unrelated things. That concentration is what the model computes.
Transition to Section 3 The story ends on her page. Section 3 of this guide is what to do when the room asks — and they will — "so what do we do?"
Section 03

The turn · From her page to their estate

Goal: Hand off from the parable to the AI Sovereignty Model — show that it does the counting they cannot do by asking around, and that it will report what they cannot answer. Then, and only then, the answer: this section holds the only moment a product is named.

◉The six places — the enterprise version of her four pages

  • L1Hardware & Network. Chips, firmware, routing — technological sovereignty.
  • L2Software Supply Chain. Registries, packages, licences that convert mid-life.
  • L3Data. Ownership, residency, jurisdiction — the place most customers believe is the whole question.
  • L4Data Processing. Pipelines, catalog, the pipes that move it — operational sovereignty.
  • L5Model. Weights, licence, alignment, behaviour that changes between API versions.
  • L6Identity & Access. Users, services — and now agents.

◉What makes it a diagnostic, not a framework

  • 01The frame is the customer's choice — EUCS, BSI C5, SecNumCloud (tab 3, The Frame). Which rules apply decides which questions they are even required to answer, and the score moves with it.
  • 02The scoring axis is replaceability under duress: if this were compelled, withdrawn, or degraded tomorrow, how long until we run without it? The scale is Unknown · Known · Rehearsed — a plan nobody has executed scores no credit.
  • 03The law already fixes the number. EU Data Act: thirty days. The sharpest single question in the play: the law says thirty days — can you actually do it?
  • 04Unknowns are counted, not hidden — and the model can return good news. The size of the unknown count is usually the finding.
  • 0530 sourced controls, each traced to a published requirement — DORA, EU Data Act, EUCS — not to our opinion.

▶Show & do — thirty seconds, then stop

1On the parable's reveal map, walk three tiles only, left column: the same few names → where you are concentrated; stopping at three → what you cannot answer; whether anyone has ever tried → rehearsed, not merely documented. Skip the rest.
2Click "Open the AI Sovereignty Model". Do not linger. Show that it exists, that it has six places and thirty sourced controls, and that the Sovereignty Register (tab 6) generates itself from their answers. Thirty seconds, maximum.
3Ask the thirty-day question out loud and stop talking. This is the moment the conversation either becomes real or does not.

◉The answer — the only place a product is named

  • 01One platform, six answers. Teradata Autonomous Knowledge Platform presents a coherent response at every one of the six places — not a stitched collection of point solutions. Read only the answers for the two places the customer named.
  • 02Envelope of their choosing. Teradata Cloud in their region, Teradata Factory on-premise (air-gap capable), or hybrid across Teradata Fabric — same policy layer, whichever envelope.
  • 03Retain, in the story's sense. Bring your own model, embedding, vector store; pinned versions at L5. Weights and data stay obtainable on their terms — the second move, in architecture.
  • 04One Policy Layer across all six. Embedded RBAC, lineage and audit apply automatically to every access and every invocation — including agent actions at L6.

!Naming discipline — do not slip

!Per Guide D044422: first reference always Teradata Autonomous Knowledge Platform; after that, "the platform". Use Teradata Cloud, Teradata Factory, Teradata Fabric, Teradata Database. Never AI Factory, Artemis, AI Workbench, or ClearScape Analytics (as product name). In-Database references say "features included in Teradata Database."

?Ask at the turn

  1. Of the six places, which two are still question marks for you? Which one would your General Counsel name first?
  2. Which assurance frame are you actually held to — and is it the one your architecture was designed against?
  3. If this returned "you cannot answer eleven of thirty" — is that a finding you could take to your board, or one that would get buried?

⚠Anticipated objections

"This is six problems where we thought we had one."
Correct — and the point is to make that visible in one conversation instead of six separate audits. The number is finite, and the same few answers usually clear several places at once. That is what the concentration finding buys.
"This will score us badly and I will have to explain it."
It can return good news, and it is built to. What it will not do is score an unknown as a zero — unknowns are counted and reported as unknowns. That is usually the most defensible artefact in the whole exercise.
"Is this just a vendor framework with a nice story on the front?"
Fair test. Every one of the thirty controls traces to a published requirement — DORA, EU Data Act, EUCS — and the frame is theirs to pick. The story ran to its end before a product was named. That was deliberate.
"We cannot rip out everything to consolidate on one platform."
Nobody is proposing that — the story would not let us. The envelope is chosen per workload; most customers consolidate where sovereignty pressure is real, usually L3 and L4, and let the rest sit until the next refresh. Coexistence is the norm.
Transition to Section 4 You leave the room with two dates for the workshop — or you don't. Section 4 of this guide covers both.
Section 04

After the meeting · Workshop or follow-up loop

Goal: Convert the meeting into a scheduled 2-hour Sovereignty Gap Workshop with named participants. If it doesn't close on a date, run the follow-up loop back through the Executor. Either way: the parable link is the only asset you send — the model is used live, never distributed.

◉Path A — the meeting closed on a workshop date

  • 01Same day. Send a calendar hold with the parable link (EN or DE). Say: "Ask your team to walk the story before we meet — it makes the workshop twice as productive."
  • 02Day 3. Confirm participants. Insist on: CIO or CISO + DPO + General Counsel + one architecture lead — plus one skeptic. The skeptic makes the register credible internally.
  • 03Day of workshop. 30-minute self-assessment (tab 4) · 60-minute place-by-place mapping · 30-minute coexistence plan. Facilitated by a Teradata sovereignty architect.
  • 04Output. A written Sovereignty Gap Register, scored by place and by exposure, with the unknowns counted. Theirs to keep. No sales follow-up committed unless they request it.

◉Path B — the meeting did NOT close on a date

  • 01Same day. Do NOT send a "let me know when to reschedule" email. Send the parable link plus one of the six places — the one their own words flagged in the meeting.
  • 02Day 3. Return to the AI Sovereignty Executor. Regenerate Stage 2 with the follow-up variant. Include a memorable quote from the meeting itself — the Executor will weave it in.
  • 03Day 7. Send one paragraph from The Kill Switch as a quote — the count, or "securing what you hold does nothing about what holds it." No link. Force a reply.
  • 04Day 14. Route through the internal partner network. If the CISO has gone quiet, the DPO or General Counsel is usually the warmer door — the register is their artefact too.

▶Follow-up guardrails

!Name no political actors in any follow-up — the story's discipline extends to your emails. The parable link is the only asset you attach; the base talk track and this guide are internal, and the AI Sovereignty Model is used live with the customer, never sent.

?Debrief with yourself

  1. Which of the six places did the customer's language actually concentrate on? Was it the one the Executor essay predicted? If not — update the essay in your notes.
  2. Did anyone say out loud that a documented exit plan has never been executed? Note the words — that sentence is the workshop's opening slide.
  3. Which routing signal (journey = Strategy, spend = ROI, existing programme = calibration) did the customer trip, even weakly?
  4. What was the memorable phrase from this meeting that becomes the hook for the Day-3 follow-up? Write it down before you forget.

🔀Post-meeting hand-off matrix

CLOSEDPath A · 2h Sovereignty Gap Workshop scheduled. Prepare the frame question (EUCS / BSI C5 / SecNumCloud) in advance — it decides the whole scoring.
STALLEDPath B · Day-3 / Day-7 / Day-14 sequence. Cycle back through the Executor.
PIVOTCustomer signalled Strategy or ROI concerns. Route to The Ascent or The Cost of More. Loop back to Sovereignty when the pivot conversation seasons.
End of guide The AI Sovereignty play converts through the workshop, and the workshop's product is the register. Everything before it is getting the count started. Everything after is selling from a register the customer built themselves. Use the base talk track. Use this guide. Do not improvise.
The guide in one paragraph

The base talk track is the storyline. This guide is the sales choreography. Outreach → story → the turn → workshop — every step anchored on the register.

🎯 What "landed" looks like

The customer tries the count and runs out at three. They volunteer which two places concern them most. Someone says out loud that a documented exit plan has never been executed. They accept two dates for the 2-hour workshop before you leave the room.

🛑 What "did not land" looks like

The room argues about the case in Beat 2 — you over-told it. Or they nod politely and name no place. Or they ask for the deck. → Do not push. Send the parable link, book a 30-minute follow-up, loop back through the Executor.

📄 Companion assets in order

1. executor.html — persona-tuned outreach.
2. Base Talk Track EN — the storyline script.
3. AI Sovereignty Model — opened at the turn, used live.
4. Landing page — the sales-play strawman.

🛡️ Discipline you must never break

No political actors named. Matter-of-fact tone, structural framing. No product named before the turn. The model is used live, never distributed. Naming per Guide D044422.

🎙️Open base Talk Track EN 🛡️Open the AI Sovereignty Model ✉Open the Executor 🏠Sales-play landing page