RCS - Sama - Sama Launches Agentic Capture

MicrosoftAnnouncement

18/02/2025 15:15

For best results when printing this announcement, please click on link below:
https://newsfile.refinitiv.com/getnewsfile/v1/story?guid=urn:newsml:reuters.com:20250218:nRSR2598Xa&default-theme=true

RNS Number : 2598X  Sama  18 February 2025

Sama Launches Agentic Capture, a Robust Data Capture Framework for Multi-Modal
Agentic AI

SAN FRANCISCO, CA / ACCESS Newswire (https://www.accessnewswire.com/)  /
February 18, 2025 / Sama (https://pr.report/79bs) , the leader in
purpose-built, responsible enterprise AI with agile data labeling for model
development and supervised fine-tuning, today announced the launch of Agentic
Capture, a feedback framework for multi-modal agentic AI. Built on Sama's
configurable capture platform, this new framework is capable of managing the
types of multi-modal, complex use cases necessary for agentic AI across a
variety of industries. Agentic Capture, built on Sama's extensible data
labeling platform, ensures consistent, accurate annotation through the use of
high-quality rubrics and instructions. The framework's precision insights and
transparent reporting provide a deeper understanding of the AI agent's
behaviors and its impact on model goals, and offers the human in the loop
(HITL) validation necessary to effectively iterate on model performance.

AI agents are models that can act autonomously to make decisions, take actions
and learn from interactions to help solve problems or achieve specific goals.
The agentic AI market is expected to explode over the next five years, moving
from a current value of USD 5.1 billion to USD 47.1 billion
(https://pr.report/79bt)  by 2030. The success of these agents will hinge on
their ability to effectively mimic human behavior, which will require
ingesting data from multiple modalities, including text, images, audio or
video. This new functionality is a step change from today's current large
language models (LLMs) which, while useful in augmenting human capabilities,
are limited by their inability to process multiple modalities. Sama's Agentic
Capture framework makes it easier to integrate multiple types of data into the
AI model, while providing the consistent human judgement needed to effectively
utilize that data.

"Sama works with some of the leading AI model builders in the world, and that
has placed us at the forefront of innovation. While LLMs have taken center
stage for AI models, they are not without their limitations. We are in the
early stages now of seeing the impact AI agents will have on the industries
and companies we work with," said Duncan Curtis, SVP of AI product and
technology at Sama. "Agentic AI is on the precipice of exploding in
capabilities and use cases, and our new framework is designed to make it
easier to leverage data - whether text-based, images, audio or video - to
enable AI agents to autonomously complete their tasks. We're excited to
continue working closely with our partners to build out a robust framework
that meets the burgeoning needs of this technology."

Use cases for AI agents run the gamut from human-like customer service in
retail and insurance; to optimizing production in manufacturing; fraud
detection in finance; resume screening, recruitment and training in human
resources; analyzing consumer behaviors in marketing and much more. Sama's
data capture framework builds on the company's decades of experience assessing
and improving AI models.

For over 16 years, Sama's proprietary HITL approach has consistently provided
models with feedback from expert annotators, validating a model's behavior and
ensuring it is performing to standards. This feedback is woven throughout the
entire model development process, including data creation, supervised
fine-tuning, LLM optimization and ongoing model evaluation, ensuring clients
can develop models in a more responsible way. Sama's work is backed by
SamaAssure™, the industry's highest quality guarantee, which routinely
delivers a 98% first batch acceptance rate.

About Sama

Sama is a global leader in data annotation solutions for computer vision,
generative AI and large language models. Our solutions minimize the risk of
model failure and lower the total cost of ownership through an enterprise
ready ML-powered platform and SamaIQ™, actionable data insights uncovered by
proprietary algorithms and a highly skilled on-staff team of over 5,000 data
experts. 40% of FAANG companies and other major Fortune 50 enterprises,
including GM, Ford and Microsoft, trust Sama to help deliver industry-leading
ML models.

Driven by a mission to expand opportunities for underserved individuals
through the digital economy, Sama is a certified B-Corp and has helped more
than 68,000 people lift themselves out of poverty. An MIT-led Randomized
Controlled Trial has validated its training and employment program. For more
information, visit www.sama.com (https://pr.report/79bu) .

Sama Media Contact:

press@samasource.org (mailto:press@samasource.org)

LinkedIn Announcements:

Lisa:

Training AI agents is even more complex than training LLMs, which is why I'm
excited to announce the launch of Agentic Capture, a framework built for
training multi-modal AI agents.

Even in well-established procedural workflows, the actions taken by human
agents are often cryptic and cluttered with superfluous information, making
interpretation challenging. When issues arise, it's frequently unclear why
they occurred or how to address them-leading to AI agents that fail to recover
and adapt.

Agentic Capture transforms raw agent traces into interpretable steps, enabling
effective oversight and meaningful improvements in AI agent performance. The
solution includes:

·      Strategic guidance that helps organizations develop a clear
roadmap by defining the data and workflows necessary for effective feedback
and alignment.

·      A process-focused feedback system that enables continuous
improvement based on detailed performance analysis.

·      Proactive error detection that identifies model behavior gaps
early and improves model understanding.

If you're in the process of training an AI agent - let's chat!

Wendy:

I am excited to announce the launch of @Sama's Agentic Capture, a feedback
framework specifically designed for multi-modal agentic AI.

This framework addresses a critical gap in the AI industry: the ability to
effectively build, monitor, and improve AI agents while maintaining
comprehensive oversight.

The solution combines strategic guidance with a sophisticated monitoring
interface, enabling organizations to track AI processes and implement
meaningful improvements based on detailed performance analysis.

Our consultative approach, process-focused feedback system, and seamless
integration with existing infrastructure help improve gaps in model
understanding resulting in higher performance.

If you're struggling with your AI Agent's performance, reach out and I can
connect you with our team of solution engineers.

#AIAgent #Enterprise #AIStrategy

Duncan:

Product teams know the struggle: building reliable AI agents is hard, and
debugging them is even harder.

When your AI agent makes a mistake, the logs are often a maze of cryptic
traces and superfluous information. You're left wondering not just what went
wrong, but how to fix it-and how to prevent it from happening again.

Here's what makes our framework a game-changer for product teams:

·      A powerful trace analysis system that converts complex agent logs
into clear, actionable insights-making debugging intuitive rather than
overwhelming.

·      Built-in tools that allow domain experts (not just engineers)
provide meaningful feedback on agent behavior

·      Proactive error detection to identify model behavior issues
before they hit production.

·      An implementation approach that integrates seamlessly with your
existing cloud infrastructure, requiring minimal setup time.

We built this because we faced many of the same problems. If you're looking to
make your AI agent development more transparent and effective, let's talk.

#AIAgent #ProductManagement #AIStrategy

SOURCE: Sama

 

This information is provided by Reach, the non-regulatory press release distribution service of RNS, part of the London Stock Exchange. Terms and conditions relating to the use and distribution of this information may apply. For further information, please contact
rns@lseg.com (mailto:rns@lseg.com)
 or visit
www.rns.com (http://www.rns.com/)
.

RNS may use your IP address to confirm compliance with the terms and conditions, to analyse how you engage with the information contained in this communication, and to share such analysis on an anonymised basis with others as part of our commercial services. For further information about how RNS and the London Stock Exchange use the personal data you provide us, please see our
Privacy Policy (https://www.lseg.com/privacy-and-cookie-policy)
.   END  NRASFWSDIEISEEE