October 27, 2026
Cheaper computer use meets the cost of real operations
OpenAI moved capable computer use down the cost curve. GPT-6.1 Terra is positioned for high-volume visual and desktop work at $1 per million input tokens and $5 per million output tokens. OpenAI’s benchmark and cost comparisons are vendor-reported. Source
Anthropic opened its economics to public scrutiny. Its S-1 confirms a proposed Nasdaq listing under ANT, with share count and price still undetermined. The filing gives enterprise buyers a new view into concentration, infrastructure commitments, governance and continuity risk. SEC filing
Agent cost is becoming an operating discipline. Gartner says 58% of surveyed infrastructure leaders view scaling AI proofs of concept as a priority, and predicts uncapped token pricing will push many agent programs more than 50% over initial cost expectations through 2027. Coverage
Health AI showed its value in orchestration. A Mass General Brigham virtual-specialist program now reaches more than 20 hospitals and is on pace for over 16,000 consultations annually, according to Andor Health. Source
MedTech’s commercial challenge remains workflow adoption. Coreline Soft received FDA 510(k) clearance for automated aortic measurement from chest CT, but its release does not include independent clinical-outcome evidence. FDA record
Top developments
1. Terra turns computer use into a volume workflow
OpenAI released GPT-6.1 Terra across ChatGPT Work, Codex and the API. The company says the model combines visual reasoning, precise computer use and clearer writing, with standard pricing 50% below GPT-5.6 Terra and GPT-6.1 Sol. It cites proposal drafting, technical-drawing interpretation, form completion and invoice entry as representative work. OpenAI
Why it matters: The important change is less about another peak benchmark and more about which tasks become economical at scale. For commercial teams, that could include preparing account briefs from mixed documents, extracting details from site photos, moving data between customer portals and CRM, and producing service records from screenshots or forms. Buyers should validate cost per completed workflow, correction rate and permission boundaries using their own systems.
2. Anthropic’s IPO filing makes vendor durability part of the AI architecture
Anthropic publicly filed its S-1 and intends to list Class A shares on Nasdaq under ANT. The offering size and price remain open. The filing reports $16.3 billion in first-half revenue and an $80.7 billion net loss, including a $72.4 billion one-time financing-related accounting charge, according to AP’s review. Anthropic | AP | SEC
Why it matters: Enterprise model selection now includes supplier economics, compute obligations, customer concentration, governance and the ability to sustain long contracts. Procurement teams should connect those questions to portability: Can a sales or service workflow move to another model without rebuilding its data, permissions and evaluations?
3. Gartner puts a meter on agent ambition
In a survey of 272 infrastructure and operations leaders, Gartner found that 58% see scaling AI proofs of concept as an important priority over the next year. It predicts that enterprises using uncapped, token-based pricing for IT operations agents will exceed initial cost expectations by more than 50% through 2027 compared with predictable subscription or bundled models. Gartner’s recommended posture is blunt: no telemetry, no deployment. ITPro summary of Gartner research
Why it matters: Agent adoption needs a workload ledger. Track cost per qualified lead, completed service case, accepted proposal or resolved escalation, along with model calls, retries, human review and exception handling. A cheap token can still sit inside an expensive process.
AI for Sales and field execution
The next field copilot may work through the screen
Terra’s positioning matters for field sales and service because much of the work is visual and trapped in systems without clean APIs. A rep may need to read a technical drawing, compare a photo with a parts catalog, update a warranty portal, produce a quote and log the outcome. Computer use can bridge those steps. The deployment test should begin with one bounded loop, clear approval points and a replayable audit trail, especially where pricing, commitments or regulated documentation are involved. Model documentation
A customer use case ties AI to a clock, not a chatbot
An AWS customer case study describes Oriserve’s use of multilingual generative AI agents for customer engagement. In one cited deployment, telecom provider Vi triggers an immediate agent interaction when a customer requests a port code, leaving less than an hour to understand the reason and respond. Oriserve reports a 75% increase in retention for that workflow. The result is vendor-reported and is not presented as an independently audited study. AWS case study
Practical takeaway: Start with moments where delay destroys value. The best sales-agent candidates have a visible trigger, a short response window, a narrow decision set and a measurable outcome.
Tuesday focus: Health Care, Pharma and MedTech
The clinical AI result is 70% more throughput, built from routing and handoffs
Andor Health says a virtual-specialist program designed with Mass General Brigham scaled to more than 20 community and rural hospitals in six months. ThinkAndor automates specialist matching, licensing and shift checks, escalation, patient-information assembly and return of documentation to the source EHR. The program is on pace for more than 16,000 consultations a year, almost 70% higher throughput than before the collaboration. These figures are reported by Andor Health. Source
Commercial lesson: The value proposition is not an isolated clinical answer. It is the removal of coordination work around the answer. MedTech and health-tech sellers should map the complete handoff, including identity, credentialing, escalation and documentation, before presenting the AI feature.
One CT scan becomes a broader measurement platform
The FDA’s 510(k) database records AVIEW Aorta Metrics, from Coreline Soft, as substantially equivalent on October 16. The software automatically segments the aorta on chest CT, classifies anatomical zones and produces diameter measurements and views. Coreline positions the clearance as an extension of a broader chest-CT platform that also supports lung nodule and coronary-calcium workflows. FDA | Coreline Soft
Commercial lesson: A platform that extracts more value from an existing scan has a strong workflow story, but clearance alone does not establish clinical benefit, reimbursement or adoption. Field teams still need evidence for accuracy, integration effort, reading-time impact and the downstream action triggered by the measurement.
Late-Friday signal: Pharma AI deals are being built around data rights and decision loops
Pathos AI, Bristol Myers Squibb and Tempus announced a multi-year neuroscience collaboration. Tempus will build a de-identified longitudinal dataset from hundreds of thousands of patients, licensed to Pathos under a five-year agreement valued at up to $55 million. Pathos says its Foundry platform will use the data to generate and test hypotheses and inform more targeted, adaptive trials. The promised scientific and clinical effects remain prospective. Pathos | Tempus SEC filing
Commercial lesson: The durable asset is the governed loop connecting longitudinal data, domain expertise, model outputs and trial decisions. That is a more defensible offer than a general research assistant.
What to watch
Terra in production: Look for third-party evidence on task completion, retry rates and cost per completed workflow, not only token price.
Anthropic’s amended S-1: The price range, share count and final risk disclosures will clarify how public markets value capital-intensive model development.
AI budget observability: Watch for CRM and service platforms to expose workflow-level spend, human-review burden and exception rates.
Health deployment evidence: For the Mass General Brigham program, the next useful proof points are transfer avoidance, time to consult, clinician workload and patient outcomes.
RSNA 2026: Coreline plans to demonstrate its U.S. chest-CT portfolio in Chicago from November 29 to December 3. Integration and reimbursement details will matter more than another capability demo.