Contact Center RFP Checklist

The voice path decides call quality, cost and who you depend on when calls fail. Settle it before anything else.

• Can we bring our own carrier (BYOC) or SIP trunk, and is any feature unavailable when we do?
• Can the platform sit on top of our existing PBX or SBC, or must voice move to your infrastructure?
• Which SIP signalling and codecs do you support, and how do you handle registration versus IP authentication?
• Do you charge a per-minute platform fee on calls that travel over our own trunk?
• Where are your media servers, and what happens to calls in progress if one fails?
• Who signs outbound calls for STIR/SHAKEN where it applies: the carrier, the platform, or nobody?
• How is caller ID chosen per campaign, and can a number be deactivated without a support ticket?
• Do agents need an installed softphone, or does the browser work, and which browsers are supported?

If you take calls as well as make them, routing is where callers feel the platform.

• How are DIDs mapped to queues, IVR flows or AI agents, and who can change that mapping?
• What queue strategies exist (longest idle, ring all, round robin), and is skills-based routing live today or on the roadmap?
• Can callers request a callback and keep their place in the queue?
• What does a caller hear when nobody answers, and is anyone alerted about the missed call?
• Can business hours, holidays and overflow rules be set per queue without a developer?
• Can an agent take inbound and outbound calls on the same seat, and how is that billed?

Dialing mode and answering machine detection drive both productivity and complaints. Ask how each is controlled, not just whether it exists.

• Which dialing modes are included in each plan: preview, power, progressive, predictive?
• How does predictive pacing decide how many lines to dial, and how is the abandon rate held under the limit we set?
• How does answering machine detection work: beep detection, greeting analysis, or transcript classification?
• In which languages has AMD been tested, and can you share the method and the conditions of the test?
• Can AMD sensitivity be set per campaign, and what happens to a call the AMD is unsure about?
• What is the false-positive handling path: how is a live person wrongly classified as a machine detected and corrected?
• How are retry rules, attempt limits and callback scheduling configured per campaign?
• Can a supervisor pause, edit or stop a running campaign instantly?

Customers reply on the channel they prefer. The question is whether the thread and the opt-outs stay joined up.

• Which channels are native (voice, WhatsApp, SMS, email), and which depend on a third-party connector?
• Do all channels for one customer appear in one conversation history?
• Does an opt-out on one channel apply to the others automatically?
• Are WhatsApp templates, buttons and media supported, and who handles template approval?
• Is message delivery reported per message, and can a send in progress be paused or cancelled?

Ask to hear real calls, not a scripted demo. The evaluation matrix in this guide goes deeper.

• Can the AI voice agent answer inbound calls and run outbound campaigns on the same telephony as human agents?
• Can callers interrupt the agent mid-sentence (barge-in), and does it stop and listen?
• How do you measure response latency, and will you share median and p95 figures from our pilot calls?
• How is the agent kept to facts: knowledge base, approved price list, guardrails before speech?
• When the agent takes an action (booking, payment promise, cancellation), how does it confirm the action succeeded before telling the caller?
• How does handoff to a human work, and does the human see the transcript and context?
• Which languages and Arabic dialects are supported in production, not just in a demo?
• Who can switch the AI off, at what scope, and how fast?
• How is AI usage billed, and can we use our own model provider keys?

Assist tools help on the live call; QA tools help afterwards. Both depend on transcription quality in your languages.

• Is live transcription available in our languages, and with what delay?
• What does agent assist show during the call: summary, sentiment, suggested responses, knowledge articles?
• Can supervisors listen, whisper and barge in, and is that logged?
• Can we build our own QA scorecards, and can AI pre-score calls for a human to review?
• Are AI summaries written into the CRM, and can an agent edit them first?

You will be asked to prove results. Make sure the numbers you need exist before you sign.

• Which reports exist out of the box, and can we see them populated with demo data?
• Is there a real-time wallboard, and how often does it refresh?
• Can reports be exported (CSV, Excel, PDF) and scheduled by email?
• Can we access raw call records through an API or data export for our own BI tools?
• How are metrics such as connect rate, abandon rate and handle time defined, and are the definitions documented?

No platform can meet your obligations for you. What you are buying is controls that help you meet your obligations and records that prove what happened.

• Is the DNC list checked on every call attempt, or only when a list is uploaded?
• Can we import national and internal DNC lists, and record why a number was added?
• How are consent records stored (source, timestamp, proof), and is consent checked before dialing?
• Are calling hours enforced in the destination's local time, per country, with day-of-week rules?
• Can we set attempt limits per contact per period?
• Can recording be switched on or off per campaign or jurisdiction, and can it pause during payment details?
• How are recording and data retention periods set, and is deletion at the end of the period automatic?
• Is every blocked call logged with the reason, and can those decisions be exported for an audit?
• Which regulations do your default rule packs reference, and who keeps them current?

Ask for status, not adjectives. A planned certification is not a held one, and both are fine as long as the vendor says which.

• How is data encrypted in transit and at rest?
• How is one customer's data isolated from another's at the database level?
• Is there role-based access control, and can roles be scoped to a campaign, team or client?
• Is single sign-on (SAML or OIDC) available today, or only on the roadmap? Is multi-factor login enforced?
• Are audit records append-only, and who, if anyone, can edit or delete them?
• In which countries is our data stored and processed, including recordings, transcripts and AI processing?
• Which subprocessors handle our data, and is the list published?
• Which security certifications do you hold today, and which are planned, with target dates?
• Has an independent penetration test been performed? When, by whom, and will you share a summary under NDA?
• Is a data processing agreement available, and is customer data used to train shared AI models?
• How are backups taken and restore-tested, and what are your recovery objectives?

If you serve several clients, or resell the platform, isolation and branding are the product.

• Can each client be a separate tenant with its own users, numbers, lists and recordings?
• Can a supervisor be scoped to see only one client's agents and calls?
• Can each tenant have its own carrier, caller IDs and compliance settings?
• Can we brand the platform with our own name, logo and domain?
• Is there a partner console for creating tenants and seeing usage across them?
• How is billing split: can we invoice our clients ourselves?

An integration that writes data back is worth more than one that only displays it.

• Which CRMs have native integrations, and do they write call outcomes and recordings back?
• Is there a documented REST API, and can we read the documentation before signing?
• Are webhooks signed, and are failed deliveries retried?
• Is there a sandbox or test tenant for developers?
• Are there rate limits, and are they published?
• Do you integrate with workforce management (WFM) tools today, or is that on the roadmap?

The cheapest platform becomes expensive if the move stalls for three months.

• What does a typical migration plan look like, step by step, and who does each step?
• Can we pilot one campaign on one trunk before moving everything?
• Can we import contacts, DNC lists, dispositions and historical call records?
• Can we port our existing numbers, and who manages the port?
• Is training included, or is it a professional-services line item?
• If we leave, how do we export our data and recordings, and in what format?

Compare total cost at your volumes, not list price per seat. The gap between a pilot quote and a real invoice is where surprises live.

• Is pricing per seat, per minute, or both? What exactly does a seat include?
• Are carrier and telephony charges included, marked up, or billed by our own carrier?
• How is AI usage billed: per minute, per conversation, per token? Is there a monthly cap we can set?
• Which features are reserved for higher tiers (dialing modes, AMD, white-label, API)?
• Are there setup fees, minimum seat counts, or minimum commitments?
• What is the price difference between monthly and annual billing?
• Is there a free trial, and is it on the real product with our own calls?

Read the SLA document itself. A marketing page saying 'high availability' is not an SLA.

• Which plans include an uptime SLA, what is the figure, and what credit applies if it is missed?
• What are support hours and response times, and in which languages and time zones?
• Do we get a named contact or customer success manager?
• How are incidents communicated, and is there a status page?
• Is user documentation available for each role (agent, supervisor, admin)?