question-bank.md

Question bank

Options run from what good looks like down to the biggest gap. Show them in this order. Every question also accepts a typed answer: match it to the closest option, or ask again if it cannot be placed.

Each question names the dimension it informs, what good looks like, an example of good, and the 30-day move to use if it turns up as a gap. Use the example in the readout when that question drives a placement or a move, adapted to the leader's company.

Opening (every role)

J1. Where is your team on its AI GTM transformation journey?

Dimension: Systems. This is the self-placement; the placement checks in SKILL.md test it.

Shared AI layer. Several teams build and run agents on one foundation with permissions, monitoring and shared context.
AI as an action engine. Agents run defined workflows and skills when something triggers them.
AI as an answer engine. AI is connected to some systems and answers questions when someone asks.
Mostly manual. Research, list building and outreach are done by hand or in traditional tools.

Sufficient when: a typed answer can be placed on one of the four. If not, ask which of the four is closest, naming them.
Good looks like: several teams run agents on one shared foundation with permissions, monitoring and shared context.
Example: marketing's event follow-up, sales' inbound qualification and CS's renewal prep all read the same account context, and each team changes its own workflow without a ticket.
30-day move: pick one repeatable workflow and run it as a triggered agent with an owner, an approval boundary and an outcome metric.

P1. Think about the new pipeline you created in the last 6 months. Roughly what % came from each source?

Typed only. Prompt with the sources: inbound (website, forms, content, SEO), outbound (SDR or AE prospecting), LinkedIn and social, events and webinars, partners and referrals, product-led signups, existing customers. Example answer: 50% inbound, 25% outbound, 15% LinkedIn, 10% events.

Sufficient when: it names sources with a share for each and covers most of the pipeline. "100% outbound" passes; "mostly inbound" or "a mix" does not.
Use: the largest source is their dominant channel and decides which play leads the 30-day plan.

P2. What are your top priorities for the next two quarters? Pick all that apply.

Multi-select. Options by role:

Marketing: Increase pipeline / Increase revenue from marketing / Expand into a new channel (LinkedIn, events, partners) / Improve inbound conversion / Speed up lead follow-up with sales / Prove marketing's impact on revenue / Launch or scale ABM / Scale content with AI
Sales: Increase pipeline / Increase revenue and hit quota / Expand outbound (email, LinkedIn, phone) / Raise win rates / Shorten sales cycles / Ramp new reps faster / Improve forecast accuracy / Get the team using AI day to day
RevOps: Increase pipeline coverage / Trusted CRM data / Launch new channels or motions faster / Build signal routing and lead scoring / Improve forecast accuracy / Consolidate the tool stack / Roll out AI agents safely / Prove attribution
Customer success: Increase net revenue retention / Reduce churn / Grow expansion revenue / Faster onboarding and time to value / Scale CS without adding headcount / Better health scoring / Smoother sales-to-CS handoffs / Customer advocacy and referrals
Founder or CEO: Increase pipeline / Increase revenue / Expand into a new channel or market / Grow without adding headcount / Prove ROI on AI / Build a repeatable pipeline engine / Improve forecast predictability / Hit board or fundraise goals
Sufficient when: at least one is picked, or a typed priority names a concrete goal ("grow" is too vague; "grow our LinkedIn motion" passes).
Use: decides which motion gets the logic test and how the gaps are ranked.

Q10. Who in your organization can build a GTM workflow today?

Dimension: People

Anyone in GTM, inside guardrails
RevOps only
Engineering
Nobody

Sufficient when: a typed answer names who builds (a role, a team or a number of people). "It depends" is asked again.
Good looks like: authorized operators in every function build inside clear permissions, tests and approvals.
Example: a marketing ops manager builds the webinar follow-up workflow from an approved template on Monday; RevOps reviews what it can write to the CRM, and it runs on Wednesday's webinar.
30-day move: define builder roles and release one safe self-service path before widening access.

Q19. What is the one number you are measured on this quarter?

Free text. Offer two or three options that fit the role and accept their own: marketing: pipeline sourced, MQL to SQL conversion; sales: new ARR, win rate; RevOps: forecast accuracy, pipeline coverage; customer success: net revenue retention, gross churn; founder: ARR growth, burn multiple.

Good looks like: one business outcome with a baseline, a target, an owner and a period.
Example: "Marketing-sourced pipeline, $4M this quarter against $3.1M last quarter, owned by the VP Marketing."
Use: every move in the readout names this number and how it moves it.

Marketing

Q11. What is your current process for turning anonymous website visitors into meetings and pipeline?

Dimension: Systems

Identified, researched and worked automatically, with outreach sent
Identified, researched and routed with outreach drafted; the rep approves
Identified and dropped into Slack, a CRM task or a report
We don't identify anonymous visitors

Good looks like: a qualified visitor is identified, researched for fit and context, and worked right away: automatically when confidence is high, with one-click rep approval otherwise.
Example: a target account's VP visits the pricing page twice on Tuesday morning; by noon the owning AE has a brief on the account and the visitor, and a drafted note to approve.
30-day move: build one visitor-to-meeting workflow with an ICP filter, ownership rules, a response-time target and a clear line between automatic sends and rep approval.

M6. Which buying signals does your team act on today, and how?

Dimension: Data

Several signals (intent, LinkedIn engagement, events, product usage) combined per account and acted on automatically
Several signals, but someone combines them by hand
One main signal, for example form fills
We don't act on buying signals yet

Good looks like: several signals are resolved to the account, combined, and trigger the right action with the evidence attached.
Example: a webinar attendee, a LinkedIn comment from her CFO and a spike in category research all land on one account, and the account moves into this week's priority list with all three listed as the reason.
30-day move: pick three signals you already have, combine them per account, and route the accounts that cross a threshold to an owner.

M7. When marketing hands a lead to sales, how fast and how consistently does sales follow up?

Dimension: Process

Every lead, within the hour, and it is tracked
Most leads, within a day
Some leads, when reps get to them
We don't really know

Good looks like: every handed-off lead is researched, routed and followed up within the hour, and follow-up is tracked back to marketing.
Example: a lead comes in at 3:10pm; the rep sends a researched first touch at 3:35pm, and the weekly report shows 96% of leads worked within the hour, by rep.
30-day move: set a speed-to-lead target, draft the first touch automatically, and report follow-up rate by rep weekly.

Q13. Who follows up on event and webinar attendees, and how fast?

Dimension: Process

Same day, through a reusable workflow
Same week, through a manual process
Later
Nobody

Good looks like: one workflow turns every attendee list into enriched, prioritized, personalized, approved same-day follow-up.
Example: the webinar ends at 1pm; by 5pm every attendee from a target account has a note that references the question they asked, approved by their rep.
30-day move: write and test one event follow-up workflow before the next event.

M2. When a high-intent person or account does not convert, what happens next?

Dimension: Systems. Marketing default when the adaptive rules do not match and J1 is above answer engine.

Routed immediately with account context and a timed play
Alerted to an owner
Reviewed in a batch
Nothing consistent

Good looks like: unconverted demand enters a timed, owned play that keeps the signal, context and reason to follow up.
Example: a prospect starts the demo form and abandons it; within the hour the owning SDR has the account, the page they were on and a suggested opener, and a reminder fires if nothing happens by end of day.
30-day move: pick one high-intent event, set a response-time target, and measure how many qualified accounts get action in time.

M5. How consistently does AI-generated content follow your approved positioning and claims?

Dimension: Process. Marketing default when the adaptive rules do not match and J1 is answer engine or below.

Tested against a maintained brand and claims library
Templates plus human review
Depends on the person prompting
It is not controlled

Good looks like: generated content is checked against current positioning, approved claims and banned language before use.
Example: a draft claims "50% faster onboarding"; the check flags it because the approved claim is "onboarding in two weeks", and the writer fixes it before it ships.
30-day move: build a versioned messaging and claims library and test a sample of outputs against it.

Sales

Q1. How much of your sales team's week is manual work such as research, list building, CRM updates and follow-up?

Dimension: Process. Also asked on the founder track.

Under 2 hours per rep
2 to 5 hours
5 to 10 hours
Over 10 hours

Good looks like: systems assemble research, update records and prepare follow-up; reps keep judgment over customer moments.
Example: an AE opens Monday to briefs on this week's twelve meetings, CRM notes already logged from last week's calls, and follow-ups drafted; her research time drops from a day a week to under two hours.
30-day move: time one representative week by task, then automate the two highest-volume chores with an owner and an accuracy check.

S6. When someone joins your sales team, how do they learn your playbook, messaging and process?

Dimension: People

They run shared workflows that carry our best practice from day one
A maintained playbook plus shadowing
Shadowing and scattered docs
They figure it out

Good looks like: new hires work from the same maintained context (ICP, messaging, objections, playbooks) your agents use, so best practice is runnable on day one.
Example: in week one a new AE runs the account research workflow the top rep built, and her first-call briefs look like the top rep's.
30-day move: capture your best rep's research and outreach as one shared workflow and make it part of the next hire's first week.

S1. When pipeline is short, who owns creating it?

Dimension: Process

Named ownership and targets by segment and channel
A shared AE and BDR plan
Leadership asks everyone to prospect
Ownership is unclear

Good looks like: ownership, target accounts, channels and contribution targets are explicit by segment.
Example: mid-market pipeline is owned by the BDR team (60%) and AEs (40%), each with named accounts and a weekly contribution review; nobody hears "everyone prospect this week".
30-day move: publish a one-page pipeline responsibility model and review contribution by segment weekly.

S2. Can reps prioritize accounts using verified signals instead of static lists or intuition?

Dimension: Data

Ranked account actions with evidence and timing
Signal dashboards plus rep review
Static lists with occasional intent
Mostly intuition

Good looks like: reps get a ranked action queue with evidence, timing, history and a reason to act now.
Example: a rep's morning queue says "Account X: new CRO started 3 weeks ago, two visits to the integrations page, lost deal 9 months ago on timing. Act today."
30-day move: back-test three signals against recent opportunities and build one prioritized queue from the ones that predicted wins.

S3. Is there a defined boundary between work AI may perform and work an SDR or AE must own?

Dimension: People. Sales default when the adaptive rules do not match and J1 is answer engine or below.

Documented by task risk and buyer impact
Team norms with manager review
Individual rep choice
No

Good looks like: the boundary is set by risk, reversibility, relationship impact and required judgment.
Example: research and CRM logging run automatically; first-touch emails need rep approval; pricing, negotiation and anything to an open opportunity stay human.
30-day move: classify each sales step as automatic, assisted, approval-required or human-only.

S4. How do call insights and seller activity reach the CRM?

Dimension: Systems. Sales default when the adaptive rules do not match and J1 is above answer engine.

Automatically, with evidence, deduplication and validation
Automated for some activities
Rep-entered summaries
They usually do not

Good looks like: insights and activity update the CRM automatically with source, confidence and deduplication.
Example: after a discovery call the opportunity shows the new champion, the stated timeline and the competitor mentioned, each linked to the moment in the call it came from.
30-day move: send one high-value call insight to a staging step, validate it, then allow CRM writeback.

RevOps

Q5. How long does it take to launch a new GTM motion such as a segment, signal or sequence?

Dimension: Process. Also asked on the founder track.

Days
Weeks
A quarter
Longer

Good looks like: data, routing, content, approvals and measurement are modular, so a new motion launches in days.
Example: leadership decides on Monday to target companies that just raised a Series B; by Thursday the signal, routing, messaging and reporting are live, reusing parts from the inbound play.
30-day move: document the last launch, find the longest dependency, and make it a reusable component.

Q15. How much of your CRM data would you trust in a board deck without cleaning it?

Dimension: Data

All
Most
Some
Little

Good looks like: definitions, ownership, validation and exception handling are explicit, so the CRM is board-ready as is.
Example: the CFO pulls pipeline by source straight from the CRM for the board meeting, and nobody spends the week before reconciling spreadsheets.
30-day move: pick the five board-critical fields, add validation rules, and route exceptions to owners continuously.

Q6. Can your GTM infrastructure capture signals in real time, connect them at both the person and account level, and trigger the right response?

Dimension: Systems

Real-time at both levels, with automated action
Combined manually before someone acts
Captured in separate tools or at only one level
No consistent signal capture

Good looks like: signals resolve people to accounts with source and history; high-confidence combinations act automatically, the rest reach a person with the evidence.
Example: a champion at a customer moves to a target account; within a day the new account is flagged, linked to the old relationship, and the AE has a warm intro drafted.
30-day move: pick three high-value signals and define source, identity rule, threshold, next action and exception path for each.

Q7. Do marketing, sales, customer success and RevOps work from one shared GTM brain for accounts, people, ICP and decisions?

Dimension: Data

One shared GTM brain
Several connected sources
Tribal knowledge
No

Good looks like: every function reads and writes one maintained record of fit, buying group, history, signals, promises and decisions.
Example: when the CSM opens a new customer, she sees why marketing targeted them, what sales promised and who signed, without asking anyone.
30-day move: define the minimum shared account record, name owners, and connect the two systems that lose the most context.

R1. Where does signal orchestration live relative to your CRM?

Dimension: Systems. RevOps default when the adaptive rules do not match and Q15 is Most or All.

A governed layer that writes validated outcomes to the CRM
Defined workflows inside the CRM
Point tools write directly
No architecture decision

Good looks like: a governed layer resolves identity, applies policy, keeps evidence, and writes validated outcomes to the CRM.
Example: three tools report activity on one account; the layer merges them into one account update with the sources attached, instead of three tools each overwriting the lead status.
30-day move: draw today's signal-to-CRM flow and set a staging, identity and writeback policy.

R3. How do you prevent enrichment, intent and automation tools from muddying the CRM?

Dimension: Systems. RevOps default when the adaptive rules do not match and Q15 is Some or Little.

Staging, identity resolution, confidence thresholds and write rules
Field-level rules and review
Each tool writes its own fields
We clean it later

Good looks like: external data passes staging, identity resolution, confidence thresholds and controlled writeback.
Example: an enrichment tool returns a new title for a contact at 60% confidence; it waits in staging instead of overwriting the title the AE confirmed on a call last week.
30-day move: stop direct writes from one noisy source and route it through staging.

Customer success

Q22. How do you find expansion and upsell opportunities in your customer base today?

Dimension: Systems

Signals routed to CSMs
QBR reviews
CSM gut feel
We don't look

Good looks like: product, relationship and company signals are scored and routed to the CSM with evidence and timing.
Example: a customer hits 90% of their seat limit and opens two job posts for the team that uses the product; the CSM gets both, with the renewal date, the same week.
30-day move: back-test two expansion signals against recent wins and route qualified accounts to a CSM queue.

Q23. How early do you see churn risk before a renewal?

Dimension: Data

90 or more days out, from signals
At the renewal call
When they tell us
After they leave

Good looks like: risk shows 90+ days out from validated usage, stakeholder, support and sentiment signals.
Example: four months before renewal, usage drops 30% and the champion's LinkedIn shows a new job; the CSM has a risk play running before the customer says a word.
30-day move: test an early-warning set against the last ten renewals and assign a play to each risk type.

Q21. When a customer signal fires, can customer success interpret it and launch the right action with full account context?

Dimension: Data

The right play launches automatically with full context
Context is available, but someone must decide and act
The team assembles context across tools
The signal is missed or no play follows

Good looks like: usage, exec, champion and support signals are read against the whole relationship and launch an expansion or risk play with evidence.
Example: a new VP joins a customer; the CSM gets a brief with the account's history, open tickets and what was promised in the sale, plus a drafted intro, the same day.
30-day move: join one expansion signal and one churn signal to customer history, each with a play, an owner and a threshold.

Q24. Does a CSM know what sales promised before taking over an account?

Dimension: Process

Automated handoff brief
Handoff call
They read the CRM
No

Good looks like: a brief is generated at close with goals, stakeholders, promises, risks, commercials and source links.
Example: the deal closes Friday; Monday morning the CSM has a one-page brief including "promised SSO by end of Q2" with a link to the call where the AE said it.
30-day move: make a close-to-CS handoff template and generate it from the deal record and call evidence.

C1. Are renewal, churn, downgrade and expansion events represented in one consistent post-sale model?

Dimension: Data. Customer success default when the adaptive rules do not match and Q23 is "when they tell us" or "after they leave".

Yes, with reconciled ARR movement and forecasting
Mostly, with manual exceptions
Separate pipelines and spreadsheets
No consistent model

Good looks like: one event model reconciles ARR movement across CRM, billing, forecast and CSM work.
Example: finance and CS report the same net revenue retention number for last quarter without a meeting to reconcile it.
30-day move: define canonical post-sale events and reconcile a sample of recent renewals across CRM and billing.

C2. Can every customer receive a value story based on actual usage and outcomes?

Dimension: Systems. Customer success default otherwise.

Automatically, with metric gates and channel rules
Generated for priority accounts
Built manually for QBRs
No

Good looks like: usage and outcome data produce a credible, customer-specific value story through approved channels.
Example: sixty days before renewal, every customer gets a one-page summary of what they achieved with the product, built from their own usage, reviewed by the CSM.
30-day move: automate a value summary on one measurable outcome for a small cohort.

Founder or CEO

The founder track also uses Q5 and Q1 above.

Q12. What share of pipeline can you attribute to a signal or campaign with confidence?

Dimension: Data

Over 75%
50% to 75%
25% to 50%
Under 25%

Good looks like: most pipeline has source and influence evidence that survives deduplication and stage changes.
Example: the board asks what the event budget produced, and the answer is a list of opportunities with the touchpoints that created them, not a guess.
30-day move: standardize campaign, signal and opportunity IDs, then reconcile last quarter before adding attribution models.

E1. What proof is required before an AI GTM initiative receives more budget?

Dimension: Data

Measured business outcome, cost, risk and adoption
Funnel improvement with a baseline
Usage and anecdotal wins
No standard proof

Good looks like: funding needs a baseline, a business outcome, total cost, a quality bar, adoption and a risk view.
Example: the inbound agent gets budget for a second region because it raised meetings per lead from 8% to 12% against last quarter, at a known cost per meeting, with reps approving 90% of drafts unchanged.
30-day move: define one investment gate and apply it to the next AI GTM funding decision.

E4. Can leadership compare an AI investment with the human capacity and vendor spend it replaces?

Dimension: Data. Founder default when the adaptive rules do not match.

Yes, using total cost and outcome quality
A directional capacity estimate
Hours saved only
No

Good looks like: leaders compare cost, capacity, quality, speed and risk across people, vendors and AI.
Example: before opening two SDR roles, leadership maps the week of an SDR, sees that research and list building are covered by agents, and hires one SDR focused on calls instead.
30-day move: build a before-and-after capacity model for the next initiative, including the oversight work it adds.

Production readiness (one per interview, adaptive)

N1. How do you decide what company and customer data an AI GTM workflow may access?

Dimension: Systems

Policy by data class, with least privilege and audit logs
Approved tools and manual review
Team-by-team judgment
No defined rule

Good looks like: access follows data class, least privilege, purpose and retention rules, with audit logs.
Example: the renewal agent can read usage and support tickets but not billing details; security approved the data classes once, and the audit log shows every read.
30-day move: classify the data one workflow uses and remove access broader than its purpose.

N2. When a model, prompt, vendor or integration changes, how do you know your GTM workflows still work?

Dimension: Process

Automated tests and alerts
Scheduled regression checks
Someone spot-checks outputs
We find out when users complain

Good looks like: representative tests run on every relevant change, with thresholds, alerts and rollback.
Example: a model update changes how first-touch emails open; twenty saved test cases catch it the same day, before a single prospect sees one.
30-day move: build a small regression set from real cases and run it whenever a dependency changes.

N3. Who owns an AI GTM workflow after it launches?

Dimension: People

A named business owner, with technical support and service levels
RevOps or IT owns the queue
The original builder maintains it informally
No clear owner

Good looks like: a named business owner owns the outcome; a technical owner owns reliability and changes.
Example: the demand gen lead owns the inbound agent's meeting rate; RevOps owns its uptime; when the builder leaves, nothing stops.
30-day move: add an owner, a service level, a review cadence and retirement criteria to every production workflow.

N4. Can you see the cost of an AI workflow per useful business outcome?

Dimension: Data

Cost per accepted outcome and downstream result
Cost per run or per user
Total vendor spend only
No

Good looks like: cost is measured per accepted output and tied to meetings, pipeline, retention or expansion.
Example: the event follow-up workflow costs a known amount per approved email and per meeting booked, and that number sits next to the cost of an SDR doing the same.
30-day move: instrument data, vendor and review cost for one workflow and link it to an outcome.

N5. Before an AI output updates the CRM or reaches a customer, how is quality controlled?

Dimension: Process

Risk-based tests, confidence thresholds and approval gates
Rules plus human review
Occasional spot checks
No standard control

Good looks like: controls match the risk of the action: validation, confidence thresholds, approvals, logging and rollback.
Example: research notes write to the CRM automatically; emails to new prospects need rep approval; anything to an open opportunity or a customer stays human.
30-day move: classify actions by risk into automatic, reviewed and human-only.

N6. Can you connect AI-assisted GTM work to pipeline, revenue, retention or expansion?

Dimension: Data

Yes, with a defensible control or attribution method
Directionally, with linked funnel metrics
Activity and usage only
No

Good looks like: business impact is separated from activity using baselines, controls or defensible attribution.
Example: half the territories ran the new outbound workflow for a quarter and half did not; the difference in meetings and pipeline is the number that goes to the board.
30-day move: pick one outcome and set the baseline and comparison method before scaling the workflow.

N7. When an AI system recommends an account action, can the user see why?

Dimension: Data

Sources, timestamps, confidence and reasoning are attached
Source links are available
A summary without evidence
No explanation

Good looks like: each recommendation shows evidence, recency, confidence and the reason for the action.
Example: "Call Account X this week" comes with the three signals behind it, their dates and links, so the rep can disagree in ten seconds instead of ignoring it.
30-day move: require evidence and timestamps in one recommendation workflow and let users flag bad reasoning.

N8. How do you decide whether to buy, build or staff a GTM workflow?

Dimension: Process

A decision rubric covering differentiation, data, risk, economics and ownership
An architecture and ROI review
Ad hoc tool comparison
Whoever asks loudest

Good looks like: a repeatable rubric weighs maintenance, data rights, speed, risk, economics and ownership.
Example: before the GTM engineer builds a custom enrichment pipeline, the team checks who maintains it in a year, what data it needs, and what buying or hiring would cost; the answer is written down.
30-day move: run the rubric on the next workflow before any vendor trial or internal build.