GTM teams are being asked to personalize more, move faster, and scale campaigns, content, and account workflows without sacrificing quality. At the same time, enterprises are losing control of AI spend, burning through annual AI budgets in a matter of months and spending up to $500 million a month on uncapped AI.
Autonomous agents can deliver at the scale and quality enterprise teams need. But only if they run efficiently and are managed effectively.
Today, we’re launching major upgrades to WRITER built for that shift, helping GTM teams scale agentic workflows in production with greater cost-efficiency, quality, and control:
- Palmyra X6, our latest flagship model, engineered to deliver frontier-level performance on agentic GTM workflows, at a fraction of the cost.
- A faster, more reliable WRITER Agent that completes long-running, multi-step work faster and more efficiently across every model, including third-party ones.
- New governance and reporting in AI Studio, so admins can see how WRITER Agent is being used, control spend, and match the right model to the right task.
Why we built this
As agentic work scales across GTM teams, token spend becomes unpredictable. Every workflow, playbook, and model choice adds up. Without a governed way to match model cost to task complexity, teams end up in one of two places: overspending on expensive models for simple work, or creating model sprawl trying to avoid it.
The teams getting real value from agents are the ones running them in production, across thousands of employees, on the work that actually moves revenue. That takes a model built for efficiency, an agent that holds up under pressure, and controls that keep spend aligned to ROI as adoption grows.
That’s what this release delivers.
Palmyra X6: frontier performance, built for efficiency
Palmyra X6 is our new flagship model. It’s engineered for the work most teams run every day: the high-volume personalization, research, and content work where marketing and sales teams need to scale without the bill running away.
It delivers frontier-competitive performance while keeping cost and time per unit of work low, so you can deploy it across thousands of employees and keep agents running long, background jobs without spend getting out of hand.
Per-task cost and latency are roughly halved versus the previous generation, with no measured quality regressions.
Runs unattended for up to 8 hours
Palmyra X6 delivers coherent reasoning without drifting. It can run unattended on a single goal for up to 8 hours, which makes background automation viable and affordable — the kind of long-running work that used to be impractical because the cost and reliability didn’t hold up.
WRITER Agent: faster, more reliable, easier to follow
WRITER Agent is now faster, more efficient, more reliable, and easier for users to follow when tackling long-running, multi-step work, regardless of the underlying model used.
Across all WRITER and third-party models tested in research, WRITER Agent now completes tasks 44% faster and at 41% lower cost per task on average, while maintaining quality. Paired with Palmyra X6, it achieves the best combined outcomes across cost, quality, speed, and reliability: an average 52% lower cost, with 48% improvement in speed and 10% improvement in quality.
What’s new
- Adaptive reasoning based on the task. WRITER Agent dynamically adjusts its approach based on the complexity of the task, giving simple questions a direct answer and more complex work a structured plan.
- Faster execution of complex, multi-step work. WRITER Agent can execute high-volume tasks in batches or delegate them to sub-agents to get through complex work faster.
- Fewer interruptions when faced with errors. WRITER Agent can recover more reliably from issues and is better at consistently calling the right tools in the right order.
- Easy to follow at a glance. WRITER Agent clearly shows what’s in progress and what’s left, with tool execution and system details collapsed under each step so you can dive in only when you want to.
- Proactively asks for input when needed. WRITER Agent prompts users with structured questions when it needs clarification before moving forward, providing the direction needed to execute work effectively.
- Surfaces work from third-party tools. WRITER Agent visualizes actions and outputs from connected tools directly in the session, from Outlook drafts to Snowflake data tables.
Under the hood, WRITER Agent is more effective at carrying context through longer sessions, executing multi-step work more efficiently by running tasks in parallel and in batches, and self-recovering from interruptions.
New governance and reporting in AI Studio
As agentic work scales, admins need a clear picture of how it’s being used and real control over what it costs. We’re launching a major upgrade to reporting and governance in AI Studio for WRITER Agent.
A centralized view of how WRITER Agent is used
Admins get a centralized Home view that shows how WRITER Agent is being used across the business: sessions, playbook and skill runs, active users, deliverables, token usage, and trends over time. Data can be filtered by team, billing group, or seat, so admins can understand adoption, performance, and spend across the org and compare period over period.
Consumption controls
Admins can view real-time token consumption, set alerts or block consumption above set limits, and monitor adoption patterns to identify top-performing workflows. That keeps agentic work focused on ROI-aligned activity as adoption grows across the business.
Flexible, governed multi-model support
WRITER’s multi-model support now extends to WRITER Agent, giving teams a selection of popular models to work with. Admins can enable models from other providers, letting users across their managed teams choose a model at the start of a session. Specialized models can also be selected for specific capabilities, such as image generation. And beyond WRITER’s own catalog, admins can bring their own models from cloud providers including Microsoft Azure, AWS Bedrock, and NVIDIA NIM.
Why this matters
Together, these updates help GTM teams scale more complex, autonomous work with confidence and control, combining model cost-efficiency and flexibility, reliable agent execution, and governed orchestration in one production-ready platform.
The shift is from piloting agents to running them in production, across thousands of employees, with spend governed and aligned to ROI instead of left uncapped. That’s what it takes to make agentic work a real part of how GTM teams operate.
Who this is for
- Enterprise marketing, sales, and GTM teams looking to automate complex workflows without blowing through their AI budgets.
- Prospective and existing WRITER customers scaling agentic work from pilots to production.
- Org and team admins responsible for driving WRITER adoption internally and keeping spend under control.
Get started
Palmyra X6, the upgraded WRITER Agent, and the new AI Studio governance and reporting features are available starting today.
- Try WRITER Agent with Palmyra X6
- Talk to us about scaling agentic workflows
- For admins: Explore the new governance and reporting controls in AI Studio
FAQ
What is Palmyra X6?
WRITER’s flagship model, engineered for the everyday work GTM teams run at high volume and optimized to pair with WRITER Agent. It’s built for efficiency, not just top-end benchmarks.
How much does Palmyra X6 cost?
$2.00 per 1M input tokens and $8.00 per 1M output tokens, or about $3.50 blended at a typical 3:1 mix. Per-task cost drops from roughly $0.25 to $0.12 versus the previous generation, with no measured quality regressions.
How fast is the new WRITER Agent?
44% faster and 41% cheaper per task on average across all models tested. Paired with Palmyra X6: 52% lower cost, 48% faster, 10% higher quality.
Can WRITER Agent run unattended?
Yes — up to 8 hours on a single goal with Palmyra X6, with reasoning that holds up over long sessions. That’s what makes background automation practical.
How does WRITER help control AI spend?
Real-time token monitoring, alerts, and hard consumption limits, plus governed multi-model support so admins can match model cost to task complexity. All visible from a centralized admin Home view.