How Vikaas uses AI.
All of it.

1. Where AI is used in service delivery

Vikaas's engine uses artificial intelligence for several specific functions. We disclose each one.

1.1 Buyer and candidate signal scoring

AI ranks LinkedIn profiles — buyers for sales pipelines, candidates for hiring pipelines — for fit against the customer's defined ICP (Ideal Customer Profile). The scoring model evaluates profile attributes including current role, company stage, geography, tenure, career trajectory, and relevant signals. Higher-scored profiles are prioritised for outreach before lower-scored ones.

Human role: Senior operators review and adjust the scoring criteria. Final prioritisation decisions are confirmed by human operators before each outreach batch begins.

1.2 Message variant generation

AI drafts initial message variants — connection request notes, first messages, follow-up messages — in the customer's defined voice and tone, using brand voice samples provided during onboarding. Multiple variants are generated per outreach batch; human operators select, edit, and approve before any message is sent.

Human role: No AI-generated message is sent to any LinkedIn user without human operator review and explicit approval. The human calibrates for voice accuracy, cultural appropriateness, timing, and contextual nuance that the AI draft may miss.

1.3 Inbound reply intent classification

When a prospect or candidate replies to an outreach message, AI classifies the reply by intent category: genuinely interested, needs more information, not now, not relevant, or requires careful handling. This classification is used to triage and prioritise the operator's attention.

Human role:Intent classifications are reviewed by human operators before any downstream action (e.g., routing to customer's Slack channel, escalating as a hot lead). Misclassifications are corrected and used to improve the model.

1.4 Internal operational efficiency

AI assists internal Vikaas workflows that are not externally visible: profile enrichment cross-referencing, list deduplication, reply prioritisation, and reporting automation. These do not directly affect customer-facing outputs but make the service operationally scalable.

Human role: All AI-generated operational outputs are reviewed periodically for accuracy and used as inputs to human decisions, not final decisions themselves.

2. Where AI is not used (human-only zone)

Critically, AI does not act autonomously in any action that is visible to a prospect, candidate, or customer. A human operator must explicitly approve before:

• Any message is sent to any LinkedIn user
• Any candidate is added to the customer's ATS as a shortlisted candidate
• Any hot lead is escalated to the customer's Slack channel or CRM
• Any targeting parameter or ICP definition is materially changed
• Any strategy recommendation is made to the customer
• Any LinkedIn account pacing or scheduling decision that could affect account health

3. What AI does with customer data

3.1 Customer-specific calibration

Within a customer engagement, customer-provided data — brand voice samples, ICP definitions, messaging history — is used to calibrate the AI output specifically for that customer. This calibration data is isolated per customer and is not shared across customers or used to improve outputs for any other customer.

3.2 No general model training

Customer-confidential data is contractually prohibitedfrom being used to train any general-purpose AI model accessible outside Vikaas's engagement with the specific customer. This is enforced through contractual relationships with all AI providers in our stack. We carry out vendor due diligence on this point before selecting or renewing any AI provider relationship.

3.3 Aggregated and anonymised data

Vikaas may use aggregated, anonymised data derived from service delivery — from which all customer identity, ICP specifics, and personally identifiable information have been removed — to evaluate and improve service performance, benchmarking, and AI model evaluation internally. Aggregated data is not considered confidential information under our MSA.

4. AI providers and confidentiality

The specific AI providers and models used by Vikaas are not disclosed publicly, both for competitive reasons and because our stack evolves as the technology does. Under NDA, on request from enterprise procurement, we can share:

• The names of our current primary AI providers
• Their relevant compliance certifications (SOC 2, ISO 27001, EU AI Act posture)
• Our contractual data handling terms with them, including training opt-out, data residency, and breach notification requirements
• Our internal process for evaluating and approving new AI providers or model upgrades

Contact trust@vikaas.ai with a procurement NDA and your specific questions.

5. The auditable trail

Every AI-generated artifact is logged in a tamper-evident system with the following attributes:

• Timestamp of generation
• Inputs provided to the AI (target profile summary, customer voice parameters, etc.)
• Output generated by the AI
• Human reviewer identity and review timestamp
• Decision taken (approved as-is, edited, or rejected) and edit summary if applicable

This log is available for customer audit on reasonable notice, subject to confidentiality and data minimisation requirements. Standard audit log compilation: 10 business days. For enterprise engagements with right-to-audit provisions in their MSA, audit access is available on demand with 5 business days notice.

6. AI ethics commitments

We make the following commitments about how AI is used ethically in our service:

• No deception:Outreach messages are transparent communications in the customer's voice. We do not use AI to create deceptive personas, false identities, or misleading claims about who is reaching out.
• No discriminatory targeting: Our ICP definitions may not include targeting criteria based on protected characteristics (race, religion, gender, national origin, disability, age, sexual orientation) under applicable law.
• Contact respect: We maintain do-not-contact lists and honour opt-out requests from prospects and candidates. Once a prospect or candidate asks not to be contacted, they are removed from all active pipelines and do not receive further outreach.
• Transparency to recipients:Outreach messages do not claim to be sent by AI when they are presented as coming from a human's LinkedIn profile. The communication is human-approved and human-calibrated; the presentation is consistent with this.
• Annual ethics review: Our AI usage practices are reviewed annually by a senior cross-functional team. Findings and any changes are documented.

7. How we manage AI model changes

When Vikaas changes AI providers, upgrades to a new model, or makes a material change to the AI's role in service delivery, we follow this process:

• Internal testing and validation against quality benchmarks before any change affects production pipelines
• 30-day advance written notification to active customers before any material change to the AI's role
• Customer right to object: if a customer objects to a material AI change, we will attempt to accommodate their requirements. If accommodation is not possible, the customer may terminate the affected pipeline without early-termination fees.
• Post-change monitoring: performance metrics are tracked for 4 weeks post-change against pre-change baselines

8. Contact for AI-related inquiries

For procurement questions about AI usage: trust@vikaas.ai

For AI ethics concerns or complaints: trust@vikaas.ai — we respond within 2 business days.

For media or research inquiries about our AI posture: hello@vikaas.ai

This policy is reviewed and updated at minimum annually. Material changes are communicated to active customers in advance.