How Leading Companies Use AI Agents: Impacts and Strategies in 2025

Key takeaway: Enterprises are moving from isolated AI features to agentic systems that understand context, take actions across apps, and deliver measurable business outcomes—ranging from double-digit efficiency gains to multi-million-dollar revenue and savings impacts.

• Companies actively deploying AI agents span customer experience platforms (Salesforce, Five9) and real estate brokerages/platforms (RE/MAX, eXp, Compass, Anywhere).
• AI agents are being used in three dominant modes:
  • Customer-facing agents that resolve inquiries, transact, and deflect calls.
  • Employee-facing copilot/agent-assist systems that summarize, route, and execute tasks across enterprise apps.
  • Workflow and operations agents that automate back-office processes at scale (document processing, underwriting, routing, marketing).
• Quantified impacts include:
  • Salesforce customers: 50,000 automated actions/week (DIRECTV), 10-point NPS lift and 25% call reduction (Falabella), doubled case deflection (Under Armour), 84% faster resolution in specific flows (Reddit), and 275,000 conversations in five months (Telepass).
  • Five9 customers: multi-million ARR deployments; AI bookings >20% of Enterprise new logo ACV; AI revenue up 42% YoY; 25% price premium for advanced AI agents.
  • Real estate platforms: +2.5 percentage point mortgage capture and +4 point warranty attach (Anywhere); 90% target for automated document submission (Anywhere); 5x increase in agent marketing output and 62% faster creation (eXp); AI-enabled global referral routing and stronger recruiting signal (RE/MAX); end-to-end agentic workflow strategy with rising platform engagement (Compass).

• Autonomous/semi-autonomous software that:
  • Understands intent from natural language.
  • Plans and executes multi-step tasks across systems (CRM, OMS, HRIS, ticketing, EHR).
  • Uses guardrails (policies, monitoring, hallucination detection) to act safely.
• Modalities:
  • Customer-facing virtual agents (web, voice, WhatsApp).
  • Agent-assist copilots (real-time transcription, summarization, suggested actions).
  • Ops agents that automate documents, underwriting, routing, and marketing workflows.

^{Note: Maturity reflects breadth of agentic use cases, depth of integration, and measured outcomes.}

• Data backbone and integration are decisive:
  • Salesforce’s Data Cloud and Five9’s deep CRM/EHR integrations show that grounded, connected data unlock trusted action-taking and measurable outcomes.
• Channels meet customers where they are:
  • WhatsApp (Falabella), voice + digital (Five9), Slack for internal agentic ITSM (Salesforce), and Teams/Zendesk/Epic integrations illustrate multi-modal reach.
• Time-to-value can be rapid:
  • Falabella moved from idea to production in two months, suggesting that pre-integrated platforms and clear intents accelerate ROI.
• Governance and safety are table stakes:
  • Five9’s AI Trust & Governance and Salesforce’s public sector certifications (FedRAMP High) highlight the need for guardrails as agents take real actions.
• Monetization models are maturing:
  • Five9’s AI bookings share (>20%), price premium (~25%), and rising AI revenue mix show willingness to pay for advanced agent capabilities.
• Real estate platforms converge on end-to-end experiences:
  • Anywhere, Compass, eXp, and RE/MAX are using agentic tools to connect lead generation, transactions, and ancillary services—improving attach rates, productivity, and retention.

• Volume and automation
  • 50,000 actions/week executed by an internal agent (DIRECTV via Salesforce).
  • 275,000 conversations in five months (Telepass via Salesforce).
  • Path to 90% automated document submissions; 15,000 docs/day (Anywhere).
  • Near-100% automation in a cosponsored program (eXp).
• Customer experience and efficiency
  • NPS +10 to 70; −25% call volume; 70% of digital interactions via WhatsApp (Falabella via Salesforce).
  • Case deflection more than doubled; double-digit CSAT improvement (Under Armour via Salesforce).
  • Average resolution time down from 8.9 minutes to 1.4 minutes in some flows (Reddit via Salesforce).
• Productivity
  • Design time reduced from 24 to 9 minutes; per-user content output up ~5x (eXp).
  • RE/MAX Aspire participation ~60% of brokerages; improved recruitment months YoY.
  • Compass platform engagement +37% YoY (sessions per agent).
• Financial and commercial signals
  • Five9: AI revenue +42% YoY; ~10% of Enterprise subscription revenue; >20% of Enterprise new logo ACV; 25% price premium for advanced AI.
  • Five9 deals: ~$3.3M, ~$1.2M, and >$1.1M ARR examples; health system expansion from ~$2M to >$5M ARR.
  • Anywhere: +2.5pp mortgage capture; +4pp warranty attach; targeted ~$100M cost savings in 2025.

• For CX and service leaders:
  • Start with high-volume intents (WISMO, billing, appointments) and agent-assist; use grounded data and tight CRM/EHR integration to improve accuracy and actionability.
• For real estate platforms:
  • Orchestrate end-to-end journeys (lead → transaction → ancillary services) with agentic automation to improve attach, speed, and margins.
  • Invest in agent productivity tooling that compounds recruiting and retention flywheels.
• For IT and operations:
  • Standardize on a data/metadata backbone (e.g., Data Cloud) and deploy governance (guardrails, monitoring, hallucination detection) to scale safely.
  • Use collaboration surfaces (Slack/Teams) as front doors for internal agents; measure time-to-resolution, deflection, and action completion rates.
• For finance:
  • Expect AI agents to command price premiums and drive upsell/expansion (e.g., Five9); track ACV share, attach rates, and incremental EBITDA from automation.

AI agents are moving beyond pilots to become operating fabric for customer and employee experiences. Salesforce and Five9 showcase cross-industry, production-scale deployments with clear ROI, while real estate leaders (Anywhere, eXp, Compass, RE/MAX) are embedding agentic capabilities to connect fragmented workflows, amplify agent productivity, and monetize adjacent services. The most successful programs share three traits: grounded data, clear intents tied to measurable KPIs, and enterprise-grade governance—turning conversational AI into reliable, revenue-generating action systems. 🚀