From the Local Government edition of October 1, 2026
Denver advances Citibot contract amendment for optional 311 AI voicebot amid multi-year deal scrutiny
City and County of Denver / Hoodline · Municipal 311 / resident experience · Denver, Colorado, United States
- Publisher
- Hoodline
- Original publication
- October 1, 2026
- Source retrieved
- 2026-10-02
- Event date
- 2026-09-22
What happened
Hoodline’s Oct. 1, 2026 report says a Technology Services amendment before Denver City Council would add $3,889,784.94 to the Citibot Inc. agreement (total $4,339,784.94) and extend service through Oct. 1, 2031, enabling an optional 311 AI voicebot purchasable later by written notice. Voice service is priced about $550,000/year for 1 million minutes (~$2.9M of the add-on over five years); Denver’s request form estimates ~$2M/year savings. The bot would identify as a bot, answer from Denvergov.org and city-provided material without wider internet access, transfer immediately on human/live agent/person, and redirect 911 callers. A Community Planning and Development call-center proof of concept is planned for 2026; citywide voice launch targeted for 2027 without full funding secured. Existing Sunny chatbot is cited at 80–95% satisfaction. Governance & Intergovernmental Relations Committee passed the item by consent Sept. 22, 2026; full Council had not voted as of the article, amid skepticism of long multi-year vendor deals.
Why it matters
Concrete large-city 311 voice automation procurement with dollars, phased POC, human-escape design, and council process—core local-government resident-services AI, distinct from public-safety dispatch.
Evidence and measured results
Cites Denver Legistar contract figures, 9News reporting on Dunwoody comments and pricing, CBS on prior 311 hour reductions, and vendor/city claims on Sunny satisfaction and possible call deflection. Full Council vote and measured voicebot outcomes are not yet available.
Limitations and uncertainty
Amendment not yet approved by full Council. Savings (~$2M/year) and vendor ~65% deflection averages are projections. Sunny satisfaction is city/vendor-reported. Follows separate Feb 2026 ComplyAI/CivCheck permitting contract—do not conflate outcomes.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-01; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- 311 and agency call centers face volume, after-hours demand, and pressure to expand self-service without abandoning callers who need humans.
- Stakeholders
- Director of resident experience
- CIO/Technology Services
- 311 operations manager
- City Council
- Agency call-center leads (e.g., CPD)
- Discovery questions
- What share of inbound 311 calls are informational vs service-ticket creating?
- Is voice an optional add-on or mandatory first hop?
- What human-escape phrases and transfer SLAs are required?
- How will Council evaluate five-year vendor terms and savings claims?
- Value hypothesis
- A grounded 311 voicebot can deflect routine informational calls if human transfer is immediate and POC metrics beat baseline hold/handle times—savings remain unproven until after POC.
- Potential engagement
- Eight-week 311 voice readiness: baseline call taxonomy, content coverage map, POC success criteria, and Council briefing pack on contract term/risks.
- Unsupported claims
- Guaranteed $2M annual savings
- 65% deflection as Denver-specific fact
- That full Council approval or 2027 citywide funding is secured
Pre-sales engineering
Role takeaway
- Fit
- Informational 311 deflection with closed corpus; poor fit for emergency or complex adjudication without human handoff.
- Architecture and integration
- Telephony → voicebot NLU → Denvergov.org knowledge base → resolve or transfer to agency queue; separate 911 redirect message.
- Prerequisites
- Curated city website knowledge base
- Queue routing for CPD POC
- Human-escape intent library
- Call logging/retention policy
- Constraints
- Pending Council authorization
- Minutes-based pricing tiers
- Incomplete citywide funding
- Caller digital confidence variation
- Security
- Limit bot to city-provided content; minimize retained caller data; enforce vendor access controls and 911 non-handling.
- Proof of value
- Containment rate on CPD POC queues
- Median time-to-human when requested
- CSAT bot vs agent
- Wrong-answer / escalation incident rate
Delivery
Role takeaway
- Work
- Secure Council approval, configure voice POC on CPD queues, train agents on takeover, publish resident notice, evaluate citywide funding.
- Dependencies
- City Council vote
- Citibot configuration
- 311/CPD staffing
- Website content owners
- Ownership
- Technology Services / resident experience owns the channel; agencies own queue-specific answers.
- Skills and adoption
- Agents learn supervised takeover; content owners keep Denvergov.org answers current.
- Governance checkpoints
- Council contract approval
- POC exit criteria before expansion
- Accessibility review of mandatory-bot risk
- Annual savings/deflection public report
- Acceptance criteria
- Bot self-identifies and honors human-escape intents
- 911 redirect message tested
- POC metrics vs baseline documented
- Optional purchase notice process defined
- Risks
- Council rejects multi-year structure
- Callers trapped before human agents
- Savings narrative outruns measured results
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Voicebot as optional add-on to existing Citibot channel; closed knowledge base from Denvergov.org; queue-specific POC (CPD) before citywide; immediate human transfer intents; separate from 911.
Governance
Who approves, reviews and stays accountable for outcomes?
Committee consent is not final authorization; five-year extension structure may face the same multi-year contract scrutiny Denver applied to other vendors. Cities should separate optional voice SKUs, funding gates, and published deflection/savings metrics before citywide mandates.
Security and privacy
What data, permissions and controls need testing?
City says it does not collect/retain caller demographics beyond name, address, and contact details; knowledge base constrained to city-provided material. Still requires vendor DPAs, call recording/retention rules, and clear 911 handoff messaging.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Voicebot may help less web-literate callers but risks blocking people who struggle with automated voice; immediate human-escape and agent jobs redesign are essential. Prior 311 hour reductions already shifted load to Sunny.
Procurement
What should contracts, pricing and exit terms secure?
Optional voice add-on via written notice avoids a second RFP but concentrates spend in a five-year amendment; Council may demand stronger justification versus shorter competitive terms.
Operating model
Which teams own the service once it runs?
Positions AI voice as first-line informational deflection with live agents for exceptions; requires content ops and queue-owner accountability beyond the vendor contract.
Publication history
- 2026-10-01Local Government · Issue 264 resources
Stable resource ID: denver-311-citibot-voicebot-amendment-20261001