Ataccama Data Trust Summit: Day 2
The Ataccama Data Trust Summit continued into its second day with an examination of how data trust is delivered through the company’s platform and toolset – readers can catch up on day 1 here, although the discussion line flows quite naturally from this point as well.
Ataccama’s head of AI Corey Keyser, reprised his role on day 1 with a refresher on what actually constitutes a data quality rule during the day 2 opening address.
The most common example of a data rule might be a validation check (i.e. a USA telephone number has 3-3-4 structure), so the rule would have multiple inputs and parameters. It would also have IF THEN structure so for example if someone wants a loan offer but they have a really low credit ranking, then that should be flagged under the “rule logic” that is specified in Atacaama Query Language (SQL).
This (phone number check example) is then put through a test rule where test data is generated through a call to OpenAI and the rule can understand that it is working properly when error inputs are correctly classed as invalid. There is also a Description Box in the Ataccama platform that can be used to clarify what the rule does… and this English language text is generated by AI and will contain a note saying that it is created by AI and should be checked.
Jay Limburn, chief product officer at Ataccama took over this session to explain why he thinks that chief data officers are “the real change-makers” who can really impact business today.
“We are seeing data being federated throughout a business and put in the hands of semi-technical people (so-called citizen scientists)… so you don’t need to know SQL any more due to the levels of abstraction that exist now,” said Limburn, painting a picture of the modern data landscape today.
“Now also, a company doesn’t have a datacenter anymore in the back of the office, it’s all now available in the cloud. So there should be a realisaton that data trust and traceability are essential and users themselves may well be of the opinion that the data feeds they are using are not necessarily to be trusted,” said Limburn. “Businesspeople are also aware of the brand damage that can be done if they use the wrong data – it could sit very badly. If a business addresses a user as a Mr rather than a Mrs, the thinking is – what else has the company done wrong at the backend? If the wrong data is out there, then AI will be working with dirty (or shall we say incomplete) data and that’s not good.”
The path to data maturity
Limburn says that the path to data maturity today is a journey through raw, organised, governed, trusted…. accessible (publishable), usable data and then onward to data that actually has business value.
“There are so many large-scale monolithic platforms out there that provide just ‘some’ of the stack (that we define for data quality that is made up of governed, trusted, accessible and usable information)… and there are also point solutions below that do just a fraction of those functions. …. and so Ataccama is positioned as the only offering that spans all of these blocks in the data maturity timeline,” said Limburn.
The company’s key verticals (which are data hungry with big complex data problems) are mostly found in banking & financial services, insurance and manufacturing (which includes utilities and pharmaceuticals).
Canonicalised rules for drinking data
Taking over from Limburn, Jessie Smith, VP of data quality at Ataccama says that data quality at scale “sounds like quite a generic term”, but most large organisations do need to use what she calls “canonicalised rules” i.e. a rule that be used and applied across any system in any environment with any vendor’s platform, such as Snowflake, DB2, Oracle, (i.e. any of the data source companies, which also includes the ERP space).
“Data observability is now coming to the fore, but it first came forward really around the time of the whole dot com era i.e. firms would want to know that if they had deployed a web application, that it was indeed online,” said Smith. “This evolved through the period when we first started to develop ML models, because we consider them to be a living, breathing thing, so we want to know what their status is at any one time.”
She led the audience through the next era in data quality; this was the time when we realised that we needed to know about the structural state of any given data source to be able to check whether some break has happened somewhere down the data pipeline, or whether some other problem has happened.
“Most observability solutions stop at surface-level checks such as schema drift, volume anomalies and just stuff like ‘freshness’ to assess whether a table is up to date. But these checks don’t assess the deeper and more complex errors that would actually cause business harm, like negative sales values or regulatory mismatches… and this is where data quality needs to be applied today,” said Smith.
“Data quality is like testing the water running through your house pipes to make sure it’s safe to drink. Data observability is like checking the plumbing to make sure the pipes are connected and the water is flowing… and that there are no leaks. Being able to leverage both functions in one unified platform is Ataccama,” said Smith, resoundingly.
Ataccama ONE Agentic
As already referenced, by way of serendipitous convenience, the company this week announced Ataccama ONE Agentic, a new iteration of its data platform. With an emphasis on AI, by replacing manual rule-writing and cleanup with automation, the platform delivers AI-ready, trusted data faster than traditional workflows. It also enables faster reporting and shorter AI development cycles.
The work here is focused on actions to replace manual rule-writing and cleanup with automation more rapidly than traditional workflows with speedier reporting and shorter AI development cycles. The ONE AI Agent is designed to act as an autonomous data worker and apply data quality rules, detect duplicates and inconsistencies and understand context. This service runs full workflows that extend from profiling data and assigning rules to validating results. It also documents each step for review.
“The next generation of AI will be defined by systems that act on data independently, not just analyse it,” said Ataccama CPO Limburn. He suggests that “for years” data engineering teams have fought fires, fixing errors after they’ve already distorted reports or slowed down projects. He says that a reactive approach like this doesn’t work when AI is making decisions in real time.
“Ataccama ONE Agentic changes this by embedding intelligence directly into how data is governed. The ONE AI Agent doesn’t just find problems; it acts on them, ensuring data stays accurate, explainable and ready for use. It shifts the focus from managing data to trusting it, because in an AI-driven enterprise, success depends not on how much data you have but on how much you can trust,” added CPO Limburn.
Data Trust Index
He says that the Ataccama MCP Server enables copilots, large language models and custom AI workflows built inside the enterprise to access governed, explainable data from Ataccama. It exposes trusted data along with its full context, where it originated, how its quality was verified, who owns it and for what purpose it can be used, so every prediction, report and decision is built on the same reliable foundation.
Each dataset also includes a Data Trust Index, which represents a quantifiable signal of reliability that downstream systems can interpret automatically. The company says that this trust is reinforced by agentic Reference Data Management, a technology layer designed to ensures that business-critical information such as product codes, customer segments, or regulatory categories all remain consistent.
The Ataccama Data Trust Index is a quantifiable signal of reliability for datasets within the Ataccama ONE platform. It provides an automated measure of a dataset’s trustworthiness, allowing data consumers (including other systems and AI agents) to understand and interpret its reliability without manual intervention.
The index is a core feature of the Ataccama ONE Agentic platform, designed to ensure data is accurate, accessible, and ready for use in critical business decisions and AI applications.
Looking deeper at the value of data can be achieved through the Data Trust Index, functionality is brought about by looking at a number of the factors that we think about when we look at data trust that includes data lineage (including information on how data has moved through an organisation’s own system and stack from an upstream and downstream point of view), data adoptions and core aspects of data quality – i.e. does the content of any given piece of data match what we expect it to be? Does a business email relating to a customer order actually contain an order?
Living layer of trust
This whole product launch is also designed to be strengthened by continuous Data Observability, a technology layer that works to track pipelines in real time to detect and resolve issues, like as schema drift, missing values, or latency.
It is important to track and remediate these factors before they affect analytics or AI models. This creates what Ataccama calls “a living layer of trust” that keeps enterprise data accurate and ready for any AI system that depends on it. In a multi-agent world, the company further states that Ataccama MCP Server provides the trust layer that allows systems, models and intelligent agents to operate from a shared, governed foundation of truth.
DQ or die!
As day 2 of this summit drew to a close, there was plenty to think about and the key question just might be… how many chief data officers have a clearly clarified data quality (DQ) policy in their core arsenal of stack functions? As Attacama would say, it’s DQ or die!
