AtScale Enterprise AI Predictions 2023

AtScale Enterprise AI Predictions 2023

As we approach the end of 2022 and reflect on AtScale’s progress over the last 12 months, we asked four of our forward-looking AtScale executives for their thoughts on the broader data, analytics, and enterprise AI markets. Here are their predictions for the year to come.

Thoughts on economy, venture capital, and leading through downturns from Chris Lynch, CEO and Chairman of the Board for AtScale

Most significant development in 2022

We are moving into the fourth economic downcycle of my career (1990, 2000, 2008, 2022). The macroeconomic causes might be different, but the outcomes will mostly be the same. It will be a time of reckoning for over-funded, over-hyped, over-valued technology companies. 

For data and enterprise AI companies in particular, we will see continued pressure to cut costs and streamline operations. That’s largely due to continued struggles in demonstrating viable paths to generating return for investors and employees.

My predictions for 2023 are simple:

  1. Enterprise spending on technology will contract. Management teams will need to figure out how long it will take as this economic softness plays out. But smart spending on data and analytics solutions will persist as leading teams figure out how to outpace their competitors.
  2. There will be some consolidation within data, analytics, enterprise AI, and cybersecurity. In 2H 2023 the largest companies will put their cash to work and take advantage of lower valuation.
  3. A revamp of traditional compensation. I’d expect (at least I’d hope) that venture-backed technology company management and boards will re-think equity compensation for their employees in order to keep them motivated and incentivized.

My simple advice for investors and leaders is that now is not the time to panic and stop investment. It is time be selective and invest in great ideas and great people.

Thoughts on the broader Enterprise AI and Analytics market from Elif Tutuk, VP of Product for AtScale

2023 will be the year to transform metrics into business capabilities. 

Most significant development in 2022

2022 was a big year for me personally as I transitioned from an incredible 12+ year run at Qlik to AtScale. You can read about my thought process in joining the AtScale team here. From an enterprise analytics perspective, I have been fascinated to watch the industry rally around the concept of data mesh

Google search trends show a 6-8x growth in search volume around data mesh, which I think is illustrative of the interest in the concept. I am writing a blog series on the topic to share how my thoughts are evolving around it and how I see the semantic layer playing a key role in realizing the true value of decentralized data product creation.

Here are my predictions for 2023:

  1. Convergence of seemingly different enterprise analytics. These functions will accelerate in 2023 — namely business intelligence, data science, AI/ML, and decision intelligence. The widespread adoption of centralized cloud data platforms laid down the foundation for acceleration. Technologies like semantic layer will establish a virtual control plane to efficiently manage different data flows with minimal physical transformation or movement of data. 2023 will also be the year for ABI and DSML tool democratization. We will see that organizations will accept the portfolio approach to adopt Analytics BI and DSML tools, empowering the business users with their own preferences on tools.To avoid fragmented data pipelines and inconsistent metrics definitions within each tool, organizations will augment the tool democratization with a semantic layer independent but integrated with cloud data warehouses and analytics tools.
  2. Organizational interest in data mesh will move from research to piloting and early implementation. The semantic layer will play a critical role in enabling centralized governance and coordination between decentralized domain-specific data product development. Analytics engineers will focus on creating domain-specific data products. They will re-discover the power of dimensional modeling as they move beyond a basic metrics design for their business units to thinking about the logical connective tissue of governed analysis dimensions (like time, product, and geography). With this focus, we will start seeing domain-specific data products and composable analytics, where each business unit owns and creates their data products based on centrally governed components managed in a semantic layer.
  3. Governance and fast access to analytics will be a must for digital processes. Organizations face rapid changes, making timely decisions is an essential capability. Adopting self-service analytics is the mainstream method to acquiring faster insights. However, self-service analytics can be slow and lead to inaccurate values at the time of reporting due to inconsistent metrics definitions across teams, especially for digital processes and automations. Pulling insights through self-service analytics tools through various steps can be error-prone and time-consuming. In 2023, we will see the use of semantic layer, not only for BI and DSML, but also for process automations. In an API-centric way or a “headless” form, the semantic layer will be used by the downstream business applications and other data sources to consume data.Business applications, especially SaaS in the cloud ecosystem, need analytics insights to be pushed in a timely manner from the semantic layer. This must be done in a headless BI format rather than be pulled by the self-service analytics process. 2023 will transform metrics into business capabilities.

Thoughts on evolution of Data Science and AI/ML in the enterprise from Gaurav Rao, EVP and GM for AI/ML at AtScale

Most significant development in 2022

January will mark my one year since joining AtScale  — see my thoughts on why I joined here. It’s been interesting to step just outside the pure play AI/ML space after 18 years at companies, including Neural Magic and IBM (Watson).

From my new vantage point, I have been surprised that  production model deployment continues to be difficult. I think MLOps tooling has hit a saturation point in the market. We need to try some different approaches to bringing AI/ML-generated insights closer to business decision makers.

Here are my predictions for 2023

  1. Massive spending, less patience for ROI. Despite a global economic slowdown, enterprise AI spending will hit an all-time record high in 2023. AI use cases that can help reduce cost, improve operational efficiencies, drive automation, and generate revenue via new business models will be the last things cut by large enterprises looking to stay afloat during a down market. As a result, we will see a tripling down on deployment and a scale-out of enterprise AI models to deliver a real ROI and consume big data. As AutoML and MLOps tooling continues to reach a point of saturation, new techniques will arise to ensure the use of AI-generated insights.
  2. The evolution of the citizen data scientist. The ambiguous term “citizen data scientist” has been used for a few years and contributed to the proliferation of AutoML tooling. This persona will formally take shape as elite business analysts that can consume and use enterprise AI byproducts — like predictions and feature insights — with the same ease they have in accessing historical data. We will see artificial intelligence model deployment rates improve as business users infuse AI efforts in their decision making processes directly.
  3. Structured workloads will remain a high priority. It’s hard to ignore the current hype cycle around language models in AI. Bleeding edge science and technology is being introduced at a rapid paceThink transformers, generative AI like stable diffusion and ChatGPT, and the highly anticipated GPT-4. That being said, a majority of what drives present day value inside the enterprise comes from less “sexy” models forecast sales, churn, inventory, failures, and other business fundamentals. While I am a strong believer that language leads the next wave of AI, we can’t ignore the current business needs created by the proliferation of data now available within the modern data stack.

Thoughts on the Semantic Layer from Dave Mariani, CTO and Founder at AtScale

Most significant development in 2022

2023 marks 10 years since the founding of AtScale. It has been an incredible journey and I feel fortunate to see that our vision of delivering a semantic layer for data and analytics is even more relevant in the age of the cloud than it was 10 years ago. 

It has been fun to watch the rise of the semantic layer (and its little cousin, the metrics layer) over the past 12 months. I am watching other players in the market adopt the notion, including ELT players like dbt Labs, data catalogs like Collibra, and data virtualization players like Starburst and Dremio

Here are my predictions for 2023

  1. 2023 will be the year of the semantic layer. The development of an independent technology category within the modern data stack helps combine the capabilities of metrics layers, data modeling, and workflow orchestration with open integration with data and analytics governance solutions.
  2. Analytics engineers will rise as infrastructure-savvy, business-oriented technicians. They will manage pipelines and data products (leveraging the semantic layer) for data consumers, evolving from traditional data engineer and BI engineer and data modeler personas.
  3. More organizations will adopt a decentralized style of delivering analytics and data to their users. This will be done by adapting the principles of data mesh to create a hub-and-spoke approach. This will deliver data products that leverage a semantic layer for sharing data models across business domains.

Get prepared for the new possibilities in data, analytics, and enterprise AI that 2023 holds. Learn more about how data analytics creates the right foundation for enterprise AI strategy.

GigaOm Sonar Chart - semantic layers and metrics stores