Lens 01
Product and roadmap
Which modules to lead with, which to follow, what agentic AI buyers really need, and whether multi-language coverage is a wedge or a nice-to-have.
Project Looking Glass
Overview
What this study will reveal
Three lenses. Each maps to a question on the table for Looking Glass leadership now.
Lens 02
Commercialization and GTM
Who owns the budget, what the buying committee looks like, what they will pay for AI safety, how they want to be priced, and which channel they prefer.
Lens 03
Strategic positioning
Whether the frontier-lab halo helps or hurts in enterprise, whether cloud-native is good enough, the in-housing threat, and which packaging path wins — specialist platform, cloud bundle, app-vendor bundle, security-platform bundle, or in-house.
Methodology
How respondents flow through the survey
A single screening question classifies each respondent by current state. Questions about a current vendor or in-house solution are shown only where they apply. Every respondent completes the final section — the Looking Glass opportunity — where the expansion thesis is stress-tested.
Specialist software vendor
Alice, Lakera, HiddenLayer, Protect AI, CalypsoAI, F5, Prompt Security, Wiz, and other specialists. Full pipeline. Provides comparable feedback across the specialist vendor set.
Platform-bundled guardrails
Cloud-provider native (AWS, Azure, GCP), model-provider native (OpenAI, Anthropic), or data-platform native (Databricks, Snowflake, Microsoft Purview). Tests the cloud-native-good-enough thesis directly.
In-house or open-source solution
Custom build or open-source tools (NeMo Guardrails, Guardrails AI, Promptfoo, PyRIT, Llama Guard) as primary approach. Captures build motivation, team size, and the convertibility signal.
BPO partner
Services partner for content review (Accenture, TaskUs, TELUS Digital, Teleperformance, Concentrix, and others). Captures services scope, spend, and openness to platform adoption.
No formal solution
Nothing formal in place yet. Forward-looking framing on what they would adopt and when.
AI application vendor handles safety
Buyer delegates safety to their AI application vendor (Sierra, Decagon, Cresta, ElevenLabs). Tests whether the application layer is absorbing the safety budget.
Section 1
Welcome and consent
Consent and disclaimer.
Welcome to Crossover Research's study on enterprise AI Safety, Security and Trust software.
Q1.1
Thank you for your interest in participating. This study explores how enterprises evaluate, adopt, and use AI safety and security software. We are seeking input from professionals involved in the selection or implementation of AI safety solutions at their organizations.
Estimated completion 20 minutes. Honorarium USD 200, payable via digital gift card, PayPal, Venmo, ACH, and other options. All responses are anonymous and analyzed in aggregate.
- I understand and am interested in the survey
- Nevermind, I am not interested
Section 2
Qualifying the right buyers
Filters to senior enterprise buyers with real AI in production and identifies their current state — named vendor, in-house build, BPO partner, or nothing in place yet.
Which best describes the type of organization at which you work?
Q2.1
- AI consulting or professional services firm
- AI safety, security, or red-teaming software vendor
- Enterprise (in-house technology team at a non-software company)
- Financial services firm
- Foundation model lab or frontier AI lab
- Healthcare or life sciences organization
- Public sector or government agency
- Software or SaaS company building AI features
- None of the above
Which best describes your organization's current state with generative AI?
Q2.2
- One or more AI features in production, customer-facing
- One or more AI features in production, internal-only
- In active pilot, not yet in production
- Planning to deploy in next 12 months
- No active AI plans
What is your organization's approximate annual revenue?
Q2.3
- Under USD 250 million
- USD 250 million to USD 1 billion
- USD 1 billion to USD 5 billion
- USD 5 billion to USD 25 billion
- Over USD 25 billion
Which best describes how your organization currently handles AI safety, security, or red-teaming?
Q2.4
- Alice (formerly ActiveFence), CalypsoAI, F5 AI Guardrails, HiddenLayer, Lakera (Check Point), Mindgard, Patronus AI, Prompt Security, Protect AI (Palo Alto Networks), Robust Intelligence (Cisco), Wiz (AI Security Posture Management), or another specialist (please specify)
- AWS Bedrock Guardrails, Azure AI Content Safety or Prompt Shield, or Google Cloud Model Armor / Vertex AI safety filters
- OpenAI Moderation or Anthropic safety classifiers (built-in model-provider safety)
- Databricks AI Gateway, Snowflake Cortex Guard, or Microsoft Purview AI Hub / Defender for Cloud AI
- Built in-house with no third-party vendor
- Open-source tools as primary approach (Promptfoo, NeMo Guardrails, Guardrails AI, PyRIT, Llama Guard, ShieldGemma)
- BPO or services partner for content review
- The AI application vendor we use (e.g. Sierra, Decagon, Cresta, ElevenLabs) handles safety on our behalf
- Nothing formal in place yet
- I'm not sure
What is your level of involvement in your organization's decisions about AI safety and security tooling — whether for current solutions or future purchases?
Q2.5
- Primary decision-maker
- Part of the decision-making committee
- Significant input but not final decision authority
- User with limited input on decisions
- No involvement in selection or management decisions
Are you, or anyone in your immediate household, currently employed by an AI safety, AI security, content moderation, or trust and safety software or services vendor?
Q2.6
- Yes
- No
This survey takes 18 to 22 minutes. Please proceed only if you can dedicate focused time.
Q2.7
- Yes, I have time to provide quality feedback
- Not right now — I will return when I have more time
Section 3
About the respondent's organization
Builds the segmentation: AI maturity, app count, monthly inference volume, agent stage, cloud, internal-versus-external AI mix, the gateway and framework layer where buying decisions sit, and whether they have an internal AI safety team. Industry and org size are not asked — both are reliably pullable from external sources and joined post-fielding via the company name captured in Section 13.
How many distinct generative AI use cases does your organization have in production today, or expect within 12 months?
Q3.1
- Zero in production today, one to two planned
- One to two in production
- Three to five in production
- Six to ten in production
- More than ten in production
Approximately how many distinct generative AI applications, models, or agents does your organization run in production today?
Q3.2
- None in production today
- 1 to 5
- 6 to 15
- 16 to 50
- More than 50
- Not sure
Approximately how many requests, prompts, or user interactions do your production AI applications process in a typical month?
Q3.3
- Under 1 million per month
- 1 million to 10 million per month
- 10 million to 100 million per month
- 100 million to 1 billion per month
- More than 1 billion per month
- Not sure
How large is your organization's dedicated AI or ML engineering team?
Q3.4
- No dedicated AI or ML engineering team
- 1 to 5 AI or ML engineers
- 6 to 20 AI or ML engineers
- 21 to 50 AI or ML engineers
- 51 to 200 AI or ML engineers
- More than 200 AI or ML engineers
Which cloud platform hosts most of your AI workloads?
Q3.5
- AWS
- Google Cloud Platform
- Microsoft Azure
- Multi-cloud (significant use of two or more)
- On-premises or private cloud
- No primary cloud, cloud-agnostic
Has your organization built, or is it actively building, an internal AI safety, AI security, or AI red-teaming team?
Q3.6
- Yes, fully built and headcount is growing
- Yes, built, headcount is flat
- In progress, hiring now
- Plan to build in the next 12 months
- No plans to build internal capability
Where is your organization on agentic AI adoption?
Q3.7
- Agents in production today, customer-facing
- Agents in production today, internal-only
- Agents in pilot, not yet in production
- Planning agent deployment in next 12 months
- Evaluating but no concrete plan
- Not on the roadmap
Which describe the AI systems your organization runs or plans to run in the next 12 months? Select all that apply.
Q3.8
- Agentic systems that call tools or APIs autonomously
- AI with access to sensitive data (PII, financial, health, IP)
- AI operating in a regulated industry
- AI used to moderate user-generated content
- Customer-facing AI features in our product
- Customer-facing chatbot or virtual agent
- Internal employee assistant or copilot
- Internal AI agents built on a low-code platform (Copilot Studio, Salesforce Agentforce, ServiceNow Now Assist, etc.)
- Processes unstructured user input (text, voice, images)
- None of the above
Of your AI and agent activity over the next 12 months, approximately what share is internal employee-facing versus external customer-facing?
Q3.9
- Almost entirely internal (90 percent or more)
- Mostly internal (60 to 90 percent)
- Roughly even (40 to 60 percent each)
- Mostly external (60 to 90 percent)
- Almost entirely external (90 percent or more)
Which AI gateways, agent frameworks, or orchestration platforms is your organization using or evaluating? Select all that apply.
Q3.10
- AWS Bedrock / Bedrock Guardrails
- Azure AI Foundry / AI Content Safety
- Google Vertex AI / Model Armor
- Databricks (Mosaic AI)
- Microsoft Copilot Studio
- Salesforce Agentforce
- ServiceNow Now Assist
- LiteLLM (open-source AI gateway)
- LangChain / LangGraph
- AWS Strands
- Nvidia NeMo
- n8n
- Internal or proprietary orchestration layer
- None or not yet decided
How long has your organization been using your primary AI safety approach?
Q3.11
- Less than 6 months
- 6 to 12 months
- 1 to 2 years
- 2 to 5 years
- More than 5 years
Section 4
How they buy
What's driving spend, what they weight in selection, who owns the budget, who else is in the room, and when the next decision lands.
Shown toAll respondents. Those with no formal solution or whose AI application vendor handles safety see forward-looking framing.
Rate how strongly each of the following is driving (or would drive) your AI safety investment priorities.
Q4.1
- Adversarial misuse by bad actors
- Agent or tool-use misbehavior
- Agents taking unauthorized actions or calling the wrong tools
- Compliance violations (EU AI Act, NIST, state laws)
- Data exfiltration through agent tool-use (agent retrieves and discloses data it should not)
- Data leakage or PII exposure through AI
- Hallucinations or inaccurate output reaching users
- Harmful content reaching users
- Model drift or behavior change in production
- Prompt injection and jailbreaks
What is driving — or would drive — your investment in AI safety and security tooling? Select all that apply.
Q4.2
- Audit finding or near-miss incident
- Board, executive, or risk committee mandate
- Cost reduction in content moderation or T&S operations
- Customer or partner contractual requirement
- Inadequacy of current cloud-native or open-source tools
- Insurance or cyber underwriting requirement
- New customer-facing or regulated AI deployment
- Preparing for agentic AI deployment
- Regulatory pressure (EU AI Act, NIST, state laws)
- Specific security incident
What solution, if any, was in place before adopting your primary AI safety approach?
Q4.3
- Alice (formerly ActiveFence)
- AWS Bedrock Guardrails
- Azure AI Content Safety or Prompt Shield
- BPO or services partner
- CalypsoAI
- Cloud-native guardrails
- Google Cloud Model Armor
- HiddenLayer
- In-house custom-built solution
- Lakera
- Manual processes
- Nothing in place
- Open-source tools
- Prompt Security
- Protect AI
- Robust Intelligence
- I'm not sure
How important are (or would be) each of the following factors when selecting AI safety tooling?
Q4.4
- Automated remediation workflows (ticketing integration, fix recommendations, auto-fix)
- Coverage of agentic AI and tool-use scenarios
- Compliance reporting and audit trail
- Custom policy authoring and policy-as-code
- Detection accuracy and low false-positive rate
- Ease of integration and time to value
- Pre-packaged industry policies (financial services, healthcare, kids and entertainment, etc.)
- Runtime latency added per request
- Speed to deploy a new custom policy or harm classifier (hours, days, weeks)
- Threat intelligence freshness
How would you rate your implementation experience with your primary AI safety approach?
Q4.5
At the moment of procurement, what was the single biggest factor that tipped your decision toward your primary AI safety approach?
Q4.6
Open-ended response
When do you expect to make your next AI safety or security purchasing decision?
Q4.7
- Already underway
- Next 3 months
- 3 to 6 months
- 6 to 12 months
- 12 plus months
- No timeline
Who owns the budget for AI safety and security in your organization?
Q4.8
- Business unit or product line
- Chief AI Officer, Chief Data Officer, or AI/ML organization
- CISO or security organization
- CTO or CIO
- Risk, compliance, or legal organization
- Shared, no single owner
- Trust and Safety organization
- Not yet established
How many people are typically involved in an AI safety or security buying decision at your organization?
Q4.9
- Just me or a small team (1 to 2)
- 3 to 5 people
- 6 to 10 people
- More than 10 people
How important is multi-language coverage in an AI safety and security solution?
Q4.10
- Critical — we serve customers in 10 or more languages
- Important — we need solid coverage in 3 to 5 languages
- Useful — primary language plus one or two is sufficient
- Not important — English-only is fine for our use case
What is the single biggest blocker today to deploying AI more broadly at your organization?
Q4.11
Open-ended response
Section 5
What they have in place today
Maps the eleven AI safety functions to who handles each today — current vendor, other software, cloud-native, manual or BPO, or nothing. Direct test of whether cloud-native is good enough.
For each AI safety function, how is it primarily handled at your organization today?
Q5.1
| Function | Primary vendor | Other software | Cloud-native | Manual / BPO | Nothing |
|---|---|---|---|---|---|
| Pre-launch adversarial testing and red-teaming | |||||
| Runtime input filtering (prompt guardrails) | |||||
| Runtime output filtering (response guardrails) | |||||
| Post-launch monitoring and drift detection | |||||
| Threat intelligence on emerging adversarial tactics | |||||
| Agent and tool-use observability | |||||
| Content moderation (user-generated and AI-generated) | |||||
| Compliance reporting and audit trail | |||||
| PII redaction and DLP for AI workflows | |||||
| Custom policy authoring and policy-as-code | |||||
| Image and multimodal content safety |
For each function where you indicated 'Other software,' please list the vendor or tool name.
Q5.2
Open-ended response
How would you rate the investment priority of AI safety and security within your organization's overall technology budget?
Q5.3
You rated AI safety's investment priority as a piped score. What factors influence this level of priority in your organization?
Q5.4
Open-ended response
For each statement about cloud-native AI safety tooling (AWS Bedrock Guardrails, Azure Content Safety, Google Model Armor), indicate your level of agreement.
Q5.5
| Statement | Strongly disagree | Disagree | Neutral | Agree | Strongly agree |
|---|---|---|---|---|---|
| Cloud-native guardrails are sufficient for our current AI safety needs | |||||
| Cloud-native guardrails fall behind on emerging threats like new jailbreak patterns | |||||
| Cloud-native guardrails lock us into a single cloud provider | |||||
| Cloud-native guardrails are good enough for non-regulated use cases | |||||
| We would prefer a specialist vendor as our cloud AI workloads grow | |||||
| Cloud-native guardrails are sufficient for agentic AI and tool-use scenarios |
Section 6
Where they want to invest next
Status and forward demand across eleven AI safety capabilities — what they use today, what they plan to adopt, and the soft middle where messaging matters.
What is your status on each AI safety capability today?
Q6.1
| Capability | Use today | Plan to adopt | Interested | Not interested | Not sure |
|---|---|---|---|---|---|
| Agentic AI safety and tool-call validation | |||||
| Content moderation (user-generated and AI-generated) | |||||
| Image and video generation safety | |||||
| Multi-agent coordination safety | |||||
| Post-launch monitoring and drift detection | |||||
| Pre-launch adversarial testing and red-teaming | |||||
| Runtime input and output filtering (AI firewall) | |||||
| Sector-specific compliance modules (financial services, healthcare, public sector) | |||||
| Threat intelligence on adversarial tactics | |||||
| Unified AI policy plane / AI gateway | |||||
| Voice AI safety |
For each capability where you indicated 'Interested' or 'Not interested,' rate the likelihood your organization will adopt it in the next 12 months.
Q6.2
- Agentic AI safety and tool-call validation
- Content moderation (user-generated and AI-generated)
- Image and video generation safety
- Multi-agent coordination safety
- Post-launch monitoring and drift detection
- Pre-launch adversarial testing and red-teaming
- Runtime input and output filtering (AI firewall)
- Sector-specific compliance modules (financial services, healthcare, public sector)
- Threat intelligence on adversarial tactics
- Unified AI policy plane / AI gateway
- Voice AI safety
Section 7
How their current solution performs
For respondents already using a named vendor, cloud-native, or open-source: support, satisfaction, mission criticality, and a per-capability scorecard on the incumbent.
Shown toRespondents who use a specialist vendor or platform-bundled guardrails as their primary solution.
How would you rate the quality and responsiveness of support provided by your primary AI safety approach?
Q7.1
Overall, how satisfied are you with your primary AI safety approach?
Q7.2
Rate your primary AI safety approach on each of the following product capabilities.
Q7.3
- Agentic AI and tool-use coverage
- Compliance and audit reporting
- Documentation and developer experience
- Post-launch monitoring and drift detection breadth
- Pre-launch red-teaming depth and coverage
- Runtime detection accuracy
- Runtime guardrail latency
- Threat intelligence currency and quality
How mission-critical is your primary AI safety approach to your organization's operations?
Q7.4
You characterized your primary approach's mission criticality as a piped score. Please explain.
Q7.5
Open-ended response
Section 8
Who they consider — and where they think safety belongs in the stack
Awareness and consideration across the full vendor set — and whether enterprises expect the AI application vendor (Sierra, Decagon, Cresta, ElevenLabs) to own the safety layer.
For each of the following vendors, indicate your status.
Q8.1
| Vendor | Not aware | Aware only | Evaluated | Trialed | Customer |
|---|---|---|---|---|---|
| Alice (formerly ActiveFence) | |||||
| AWS Bedrock Guardrails | |||||
| Azure AI Content Safety or Prompt Shield | |||||
| CalypsoAI | |||||
| F5 AI Guardrails | |||||
| Google Cloud Model Armor | |||||
| Guardrails AI | |||||
| HiddenLayer | |||||
| Lakera (Check Point) | |||||
| Microsoft Defender for Cloud AI / Purview AI Hub | |||||
| NeMo Guardrails (NVIDIA) | |||||
| Prompt Security | |||||
| Promptfoo | |||||
| Protect AI (Palo Alto Networks) | |||||
| Robust Intelligence (Cisco) | |||||
| Wiz (AI Security Posture Management) |
If you use a customer-facing AI application vendor (Sierra, Decagon, SoundHound, Cresta, Regal, ElevenLabs, etc.), do you expect that vendor to provide AI safety and security as part of its product?
Q8.2
- Yes, that should be their responsibility
- Partially, but I want my own guardrails layered on top
- No, I want to own the safety layer independently of any application vendor
- Not applicable, we don't use AI application vendors
How does your primary AI safety approach compare to alternative AI safety solutions you have used or evaluated?
Q8.3
You rated your primary approach's comparative performance as a piped score. Please explain your rating.
Q8.4
Open-ended response
Section 9
What they spend
Current BPO and services spend that could shift to platform, plus forward AI safety budget envelope.
What does your organization currently spend annually on content moderation, trust and safety operations, or related BPO and services?
Q9.1
- Not applicable, no spend
- Under USD 500k
- USD 500k to USD 2 million
- USD 2 million to USD 10 million
- Over USD 10 million
- Prefer not to say
What is your expected annual budget for AI safety, security, and red-teaming over the next 12 months?
Q9.2
- No dedicated budget
- Under USD 50k
- USD 50k to USD 250k
- USD 250k to USD 1 million
- USD 1 million to USD 5 million
- Over USD 5 million
Section 10
How they want to pay
ROI rating, what would unlock more spend, pricing structure preference, premium they would pay for agent coverage, channel preference, unified-platform vs piecemeal-stack elasticity, and whether internal employee-facing AI and external customer-facing AI are bought as one budget or two.
How would you rate the ROI from your primary AI safety approach?
Q10.1
I would be willing to pay more for your primary AI safety approach if it introduced _______. Please explain.
Q10.2
Open-ended response
Which pricing structure do you prefer for AI safety and security tooling?
Q10.3
- Custom enterprise agreement
- Flat annual platform fee
- Per protected AI application or per model
- Per-request or per-token consumption
- Per-seat or per-developer pricing
- Tiered by feature with usage caps
How much more would you be willing to pay for an AI safety solution that explicitly covers agentic AI versus one that covers only chat and generation use cases?
Q10.4
- Nothing extra, it should be included
- Up to 10 percent more
- 10 to 25 percent more
- 25 to 50 percent more
- More than 50 percent more
- Don't know
Through which channel would you most likely purchase an AI safety and security platform?
Q10.5
- AWS, Azure, or GCP marketplace
- Direct from the vendor
- Through a managed security services provider (MSSP)
- Through a systems integrator or consultancy
- Through our incumbent enterprise security vendor (Palo Alto, Cisco, Check Point, F5, CrowdStrike, etc.)
- No preference
If you were to purchase pre-launch testing, runtime guardrails, and post-launch monitoring as separate modules from different vendors rather than as a unified platform from a single vendor, how much more or less than the unified price would you expect to pay in total?
Q10.6
- I would only pay for the unified platform, not separate modules
- I would expect to pay 25 percent or more LESS in total if buying separately
- I would expect to pay roughly the same total
- I would expect to pay 25 percent or more MORE in total if buying separately
- Don't know
Do you expect to buy AI safety and security capability as one combined solution covering both internal and external AI, or as separate solutions?
Q10.7
- One combined solution covering both internal employee-facing AI and external customer-facing AI
- Separate solutions, with different vendors for internal versus external
- Only for external customer-facing AI (internal handled by IT or built-in tools)
- Only for internal AI (external handled by application vendor or not yet a priority)
- Undecided
Who owns the budget for each side at your organization — internal employee-facing AI versus external customer-facing AI?
Q10.8
Open-ended response
Section 11
How sticky their current solution is
Recommendation likelihood, replacement difficulty, renewal intent, and a verbatim on what use case they would extend their current solution into.
Shown toRespondents with a current solution. Specialist-vendor and platform-bundled users see the full set; in-house and BPO respondents see adapted prompts.
How likely are you to recommend your primary AI safety approach to a peer at another organization?
Q11.1
You rated your likelihood of recommending your primary approach as a piped score. Please explain.
Q11.2
Open-ended response
How difficult would it be to replace your primary AI safety approach?
Q11.3
How likely are you to renew your contract with your primary AI safety approach when it expires?
Q11.4
ConditionalShown only to respondents who scored 5 or lower on the renewal question above.
What are the primary factors that would cause you NOT to renew?
Q11.5
- Better alternative found
- Budget constraints
- Business needs changed
- Integration issues
- Lacks innovation or AI roadmap
- Low user adoption internally
- Missing key features
- Poor vendor support
- Reliability or uptime issues
- ROI not realized
- Other (please specify)
If you could deploy your primary AI safety approach into one additional AI use case at your company that they don't cover today, what would it be?
Q11.6
Open-ended response
Section 12
The Looking Glass opportunity
After reading the Looking Glass description: module appeal, evaluation intent, forced single-module choice, latency tolerance by use case, deployment preference, packaging appeal, agent-capability priority, Van Westendorp price points, objections, frontier-lab halo perception, and a question they would ask the CEO.
Before this survey, how familiar were you with Alice (formerly ActiveFence)?
Q12.1
- Not aware of the company
- Aware of the name only
- Aware and have read about the product
- Evaluated the product
- Trialed or piloted the product
- Current customer
When you think about AI safety, security, or red-teaming vendors, which companies come to mind first? Please list up to five.
Q12.2
Open-ended response
Based on the description, how appealing would each Alice product be for your organization?
Q12.3
- WonderBuild — pre-launch adversarial testing
- WonderFence — runtime AI firewall, sub-150ms latency
- WonderCheck — post-launch monitoring and drift detection
- Rabbit Hole — threat intelligence feed
- Agentic AI safety and tool-call validation (roadmap)
- Industry-specific policy packs (financial services, healthcare, kids and entertainment, etc.)
- A unified bundle of WonderBuild + WonderFence + WonderCheck
How likely are you to evaluate Alice for your AI safety needs in the next 12 months?
Q12.4
You rated your likelihood of evaluating Alice as a piped score. Please explain why.
Q12.5
Open-ended response
If you could only buy one of the following in the next 12 months, which would it be?
Q12.6
- A unified platform covering pre-launch testing, runtime guardrails, and post-launch monitoring
- Post-launch monitoring and ongoing red-teaming
- Pre-launch testing and red-teaming
- Runtime guardrails and runtime protection
- None of the above is a priority
For each AI safety capability, indicate the maximum acceptable latency added per request.
Q12.7
| Capability | Under 50ms | 50 to 150ms | 150 to 500ms | Over 500ms | None |
|---|---|---|---|---|---|
| Input prompt filtering | |||||
| Output filtering | |||||
| PII redaction | |||||
| Agentic tool-call validation |
What is your preferred deployment model for AI safety and security tooling?
Q12.8
- Customer VPC (private deployment within our cloud account)
- On-premises or air-gapped
- Vendor SaaS, multi-tenant
- Vendor SaaS, single-tenant or dedicated
- No preference
How appealing would each of the following packaging options be for AI safety and security capability?
Q12.9
| Packaging option | Not at all | Slightly | Moderately | Very | Extremely |
|---|---|---|---|---|---|
| Unified platform from a single specialist vendor (pre-launch + runtime + post-launch) | |||||
| Best-of-breed specialists, one vendor per capability layer | |||||
| AI safety bundled with the cloud provider running our AI workloads | |||||
| AI safety bundled with the AI application vendor we already use (Sierra, Decagon, Cresta, ElevenLabs) | |||||
| AI safety bundled with our existing enterprise security platform (Palo Alto, Cisco, Check Point, F5) | |||||
| Built and maintained in-house |
Which agent-specific safety capabilities matter most? Rank your top three.
Q12.10
- Action audit trail and forensics
- Cost and runaway-loop protection
- Cross-agent and multi-agent coordination safety
- Goal and intent verification (is the agent doing what was asked)
- Identity and entitlement propagation
- Sandboxing and blast radius controls
- Tool-call validation and authorization
Based on the Alice description, at what annual price (USD) would you consider Alice's unified WonderSuite...
Q12.11
- To be so expensive that you would not consider buying it
- Starting to get expensive, but not out of the question
- To be a bargain — a great buy for the money
- To be priced so low that you would question the quality
What, if anything, would prevent you from evaluating or buying Alice?
Q12.12
- Already committed to cloud-native or open-source approach
- Already committed to or satisfied with current vendor
- Budget constraints
- Concern about company longevity or financial backing
- Concern about coverage gaps in our use case
- Concern about data residency or privacy
- Concern about integration or implementation effort
- Frontier-lab-first positioning feels misaligned with enterprise needs
- Not familiar enough with the brand
- Other (please specify)
- Nothing significant — we are open to evaluating
Alice is best known for being the safety partner behind 7 of the 10 largest AI foundation labs and for moderating content for major consumer platforms. How does that credential affect your perception of Alice as a fit for your enterprise?
Q12.13
If you could ask Alice's CEO one question before evaluating the platform, what would it be?
Q12.14
Open-ended response
Section 13
Wrap-up
Contact info for honorarium delivery and consent for a follow-up call.
Please input your contact information so we can process your honorarium. Allow up to 7 business days.
Q13.1
- First name
- Last name
- Job title
- Company
Would you be interested in a 60-minute follow-up call to discuss your responses in more detail, for an additional honorarium?
Q13.2
- Yes, I am interested
- No, I am not interested
- Maybe, please send more detail
Thank you for your time. Your honorarium will be processed within 5 business days following completion.
Appendix
Benchmarked metrics
Nineteen metrics in this study — twelve numeric ratings, seven open-text rationales — are scored on the same scales we use across our broader research universe. Looking Glass results sit directly alongside benchmark vendors.
| Question code | Where it appears | What it captures |
|---|---|---|
| CC_RANK_Implement Ex | Section 4 (How they buy) | Implementation experience — current-solution users |
| CC_OE_Ven Selection | Section 4 (How they buy) | Why-we-chose verbatim — current-solution users |
| CC_RANK_Invest Prior | Section 5 (What they have today) | AI safety budget priority — all respondents |
| CC_OE_Invest Prior | Section 5 (What they have today) | Piped rationale on investment priority |
| CC_RANK_Support Qual | Section 7 (Current solution performance) | Support quality — specialist-vendor users only |
| CC_RANK_Customer Sat | Section 7 (Current solution performance) | Overall CSAT — specialist-vendor and platform-bundled users |
| CC_RANK_Mission Crit | Section 7 (Current solution performance) | Mission criticality — specialist-vendor and platform-bundled users |
| CC_OE_Mission Crit | Section 7 (Current solution performance) | Piped rationale on mission criticality |
| CC_RANK_Comp Compar | Section 8 (Consideration set) | Comparative rating — current-solution users, including in-house |
| CC_OE_Comp Compar | Section 8 (Consideration set) | Piped rationale on comparative rating |
| CC_RANK_ROI | Section 10 (How they want to pay) | ROI rating — all current-solution users (vendor, in-house, BPO) |
| CC_OE_Feature Enhanc | Section 10 (How they want to pay) | WTP-for-enhancement verbatim |
| CC_RANK_Recommending | Section 11 (Stickiness) | Likelihood-to-recommend — current-solution users, including in-house |
| CC_OE_Recommending | Section 11 (Stickiness) | Piped rationale on recommendation |
| CC_RANK_Replace Diff | Section 11 (Stickiness) | Replacement difficulty — all current-solution users (vendor, in-house, BPO) |
| CC_RANK_Renewal Int | Section 11 (Stickiness) | Renewal intent — specialist-vendor users only |
| CC_RANK_Alice Intent | Section 12 (Looking Glass opportunity) | Post-exposure Looking Glass evaluation intent — all respondents |
| CC_OE_Alice Intent | Section 12 (Looking Glass opportunity) | Piped rationale on Looking Glass intent |
| CC_RANK_Frontier Halo | Section 12 (Looking Glass opportunity) | Frontier-lab credential perception — all respondents |