Exscientia : 2023 Annual Report and Financial Statements

TABLE OF CONTENTS
	Page
General Information	1
A Letter from Our Interim CEO	2
Strategic Report	4
Directors' Remuneration Report	44
Directors' Report	74
Independent Auditors' Report to the Members of Exscientia plc	78
Consolidated Statement of Comprehensive Income for the Year Ended 31 December 2023	86
Consolidated Statement of Financial Position as at 31 December 2023	87
Parent Company Statement of Financial Position as at 31 December 2023	89
Consolidated Statement of Changes in Equity for the Year Ended 31 December 2023	90
Parent Company Statement of Changes in Equity for the Year Ended 31 December 2023	91
Consolidated Statement of Cash Flows for the Year Ended 31 December 2023	92
Notes to the Financial Statements	94

About the Cover:

Researchers at Exscientia analyse tissue samples at single cell resolution, collecting phenotypic and morphological data to determine treatment effect and differences in each patient. Our patient-first precision medicine platform aims to reset how we bring novel therapies to patients. By pioneering new approaches to AI-led drug design and development, we believe the best ideas of science can rapidly translate to the best medicines for patients.

General Information

Directors

Elizabeth Crain

Ben Taylor

David Hallett

Franziska Michor

Robert Ghenchev

Mario Polywka

Company Secretary

Daniel Ireland

Registered Office

The Schrödinger Building

Oxford Science Park

Oxford, Oxfordshire

OX4 4GE

United Kingdom

Company Number

13483814

Independent Statutory Auditors

PricewaterhouseCoopers LLP

3 Forbury Place

23 Forbury Road

Reading RG1 3JH

United Kingdom

Solicitors

Cooley (U.K.) LLP

22 Bishopsgate London EC2N 4BQ

1

A Letter from Our Interim CEO

Dear Shareholder,

As I enter the 5th year of my personal journey at Exscientia, I look back at our achievements in 2023 with pride.

In 2023, we increased the number of Exscientia-designed molecules in active clinical trials to four. EXS4318, our PKC- theta inhibitor, currently in development with our partner Bristol Myers Squibb, entered a healthy volunteer study. In the summer, Exscientia began enrolment into ELUCIDATE, our Phase 1/2 trial with our co-owned CDK7 inhibitor (GTAEXS617). In addition, Sumitomo Pharma advanced the clinical development of two bispecific compounds we designed for psychiatric diseases, DSP-2342 and DSP-0038.

Building a sustainable pipeline is also a necessary part of Exscientia's long-term growth strategy, and to this end we continued to advance our wholly owned LSD1 and MALT1 small molecule assets through IND-enabling studies whilst continuing to precision design the next wave of small molecules to treat a range of oncology indications.

If we truly wish to transform the way the world invents and develops new medicines, then we must also influence the pharmaceutical R&D ecosystem from within and maintain focus on our partnerships, which remain a key component of Exscientia's business strategy. We were delighted to add a new partner to the Exscientia family with the announcement of an AI drug discovery collaboration with Merck KGaA, Darmstadt, Germany in 2023. With one of our existing partners, Sanofi, we not only achieved our first preclinical milestone, but we also introduced an Exscientia-originated program into that collaboration with additional downstream economics.

As I look forward with optimism, I am also reminded of the immense challenges faced by our industry and why Exscientia exists. Patients demand better medicines that are designed, developed, and brought to market in an economically sustainable way. Unfortunately, failure continues to be the default setting in our industry. More than 90% of potential new drugs that enter clinical development do not reach approval. When a new medicine does make it to market, current estimates suggest that the attrition weighted investment made to bring that drug to approval is more than $2.0 billion and constantly rising. This is unacceptable and not sustainable. Exscientia aims to be at the heart of an ecosystem that transforms the way the world designs and develops new medicines. We will only achieve this and outperform the industry by continuing to encode and automate the entire drug discovery process.

2

The main reasons for failure in our industry are well documented and our investments are aligned to address these:

Every drug candidate that enters an oncology clinical trial will have demonstrated effectiveness in live animal models - yet only 4% of those drug candidates will make it to market. We know that patients are the best models. Building on our award winning EXALT-1 clinical study in haemato-oncology patients, in 2023 we initiated EXCYTE-1, a prospective observational study in ovarian cancer with the clear intent to extend the utility of our precision medicine capabilities into solid tumours. In addition, we also placed our clinically validated functional drug evaluation platform within Charité - Universitätsmedizin Berlin, one of the largest university hospitals in Europe. This is a clear example of Exscientia's commitment to putting patients at the centre of everything we do;

The majority of properties associated with a drug candidate are irrevocably set on the day it is first designed, but many characteristics only reveal themselves after millions of dollars have been invested in clinical development. We pride ourselves on the insights of our scientists to fully define the problems we intend to address at the very start of every project. Our leading generative AI design platform has evolved and our pipeline is a physical demonstration of our potential to design well balanced molecules and to overcome drug design problems that have been historically challenging;

A central tenet at Exscientia is that novel drug discovery is a learning problem based on the initial availability of sparse data sets. The complexities of the process simply do not align with ultra-high- throughput screening, whether that be computational or experimental. With that in mind, we have continued to invest in internal experimental capabilities to generate proprietary, high-quality and high- fidelity data to support our predictive systems and to support our learning environment. In 2023, we completed a multi-year construction campaign and opened a cutting-edge automation studio in Milton Park, near Oxford. For the first time, we are now able to combine synthesis sympathetic generative design with a highly orchestrated make and test environment that can generate thousands of data points on a bespoke set of AI designed compounds (rather than the one data point on thousands of compounds). We are already starting to see the long-term potential of this platform through significant efficiency gains in automated assay development and testing and I look forward with anticipation to realising the full potential of this capability.

We remain steadfast in our mission and our commitment to Exscientia's unique approach. We leverage and operate at the interfaces of human ingenuity, AI, automation and physical engineering. We do this because patients deserve access to an abundance of affordable, transformative drugs.

David Hallett, Ph.D.

Interim Chief Executive Officer

11 April 2024

3

Exscientia plc | 2023 Annual Report and Financial Statements

Strategic Report

Introduction

Exscientia plc (the "Parent Company") on behalf of itself and its subsidiaries (which together may be referred to as the "Group", "Exscientia", "we", "us" or "our"), is required to produce a strategic report complying with the requirements of the Companies Act 2006 (Strategic Report and Directors' Report) Regulations 2013 and the Companies (Miscellaneous Reporting) Regulations 2018 for the year ended 31 December 2023. Exscientia also filed with the U.S. Securities and Exchange Commission (the "SEC") its Annual Report on Form 20-F for the year ended 31 December 2023, which contains additional disclosures regarding certain of the matters discussed in this report.

The Parent Company was incorporated on 29 June 2021. Since 1 October 2021, the Parent Company has had American Depositary Shares representing its ordinary shares ("ADSs") traded on the Nasdaq Global Select Market ("Nasdaq") in the United States. The consolidated comparative financial statements are for the Group as a whole.

Business Overview

Overview

We are a drug design company using artificial intelligence, or AI, and other technologies to efficiently design and develop differentiated medicines for diseases with high unmet patient need. The focus of our platform is to improve the probability of successful drug development by identifying and resolving likely points of failure using our AI design technology, translational systems and clinical modelling. We have demonstrated our platform can achieve design goals beyond current industry standards by advancing multiple development candidates with differentiated properties, four of which are currently in clinical trials. Our internal pipeline is primarily focused on oncology, but we also use our design capabilities with partners to expand our pipeline, generate income and improve our technology platform.

We believe many drug candidates fail due to predictable drug design issues. For more than a decade, we have been utilising AI to overcome these design issues and create better quality drug candidates. We also integrate novel experimental and automation systems in order to test and validate our AI-based simulations. Our closed loop of virtual design and physical experimentation is a critical advantage of our company because it allows our platform to learn quickly, to generate data that would not be available externally, and to be cost effective and reproducible.

Our technology platform spans generative AI, active learning, machine learning, physics-based systems, large language models and many other predictive systems. However, the output of our technology is always a measurable drug. We have over 20 drug programmes advancing, including at least two with expected clinical milestones in 2024. Each drug we create needs to have a meaningful design advantage over known competitors that is expected to have clinical benefit and can be clearly measured.

Our lead internal candidate, a CDK7 inhibitor known as GTAEXS617 ('617), is currently in a Phase 1/2 trial with initial data expected in the second half of this year. '617 was precision designed to manage the potential toxicities associated with CDK7 and to optimise pharmacokinetics for maximal on-target efficacy.

4

Exscientia plc | 2023 Annual Report and Financial Statements

Strategic Report

We designed a PKC-theta inhibitor for Bristol Myers Squibb that they then in-licensed and are currently testing in Phase 1 clinical trials. Despite PKC-theta being a target of high interest, with more than a dozen companies having attempted to design compounds for the target, no competitor drug candidates have been sufficiently potent and selective and our candidate has the potential to be first-in-class. We have ongoing milestones and royalties associated with the programme.

We have two additional internal programmes in IND-enabling studies, respectively targeting LSD1 and MALT1. Both of these candidates were designed to mitigate known toxicities that have been seen in competitive programmes. By understanding the origin of these toxicities and designing against them we believe we have produced two candidates that will have an improved probability of success in clinical development.

Over time, we believe our transformational way of designing and developing drugs can change the industry's underlying pharmacoeconomic model, what we call 'shifting the curve'. We aim to demonstrate that it is simultaneously possible to improve probability of success through designing better quality drugs while also reducing investment requirements through improved technologies and process.

Our Strategy

Our strategy is to combine precision design, the deliberate placement of each atom in a compound, with integrated experiment, the ability to embed experimentation into our technology platform. This will enable us to design and develop better quality, balanced and differentiated medicines for patients with a higher probability of success in a faster and more efficient manner than industry average. By primarily focusing on the design of small molecules, we believe we can overcome complex issues that have impacted the success of other medicines. Our average time from initiating novel designs to first synthesis of the eventual drug candidate is approximately one year, and we typically synthesise fewer than a tenth of the number of compounds compared to conventional approaches.

Our approach aims to modernise the process of discovering and developing drugs, replacing the sequential, artisanal approach that currently dominates the industry, with an integrated, AI-first,patient-based learning system that is suited to the complexity of drug discovery.

We believe drugs often fail at the first step: design. By using precision design and integrated experiment, we believe we can accelerate the discovery of medicines and improve the probability of clinical success. We are driven to codify and optimise drug discovery, to move away from traditional sequential design, and instead leverage AI-basedmulti-parameter optimisation to scale the creation of precision engineered drugs.

Our innovative and advanced technologies and automation are designed to engineer better molecules than what has been accomplished with traditional methods. We believe our platform has the potential to fundamentally improve the probability of clinical and commercial success.

5

Exscientia plc | 2023 Annual Report and Financial Statements

Strategic Report

Precision Design

A drug's potential utility is encoded into its chemical structure from the moment it is first designed. Before a compound is ever synthesised and tested, the placement of each atom and bond will have predetermined how it will interact with the incredible complexity of human biology and disease. The molecular structure of the compound determines its potency, selectivity, safety, absorption, dose requirements and manufacturability as well as many other features that define a drug product. We believe every drug candidate should be designed at the atomic level to drive optimal efficacy with minimal side effects.

Design from any data. High quality drugs need to satisfy dozens of diverse parameters, defined as a target product profile, or TPP. No single data type, such as a protein structure or cellular assay, can inform all of the parameters necessary to design a TPP. Our AI platform is data-agnostic, capable of modelling and exploiting diverse data types, including protein structures, high content screening, pharmacology and other data, through thousands of machine learning, physics-based and other predictive models. We have also developed proprietary tech-enabled laboratory capabilities to generate a wide variety of high-fidelity screening data (high content, biophysical, pharmacological and biochemical) and structural biology data to provide differentiated insights for our projects.

Our tech-drivendesign cycle. Our design philosophy is that every molecule should be designed by an algorithm. We unlock the creativity of AI through the use of reinforcement learning, deep learning and evolutionary algorithms to precisely design and select novel compounds that meet our design objectives. Our design cycle is as follows:

6

Exscientia plc | 2023 Annual Report and Financial Statements

Strategic Report

Benefits of multi-parameteroptimisation. Conventional drug discovery approaches focus on sequential design improvements, usually starting with target potency, then selectivity, then refining other properties, which often leads to suboptimal molecules. AI excels at multi-parameter optimisation, and our platform can design against more complex endpoints than have been conventionally possible. We have successfully designed molecules with very little starting data, with and without x-ray structure, and directly using high dimensional, high content data. We can also design small molecule bispecifics and drugs with complex phenotypic endpoints. We are not aware of any other design system that can incorporate such a breadth of data types into design.

Drug design is a learning problem. When designing truly innovative drugs, there will be insufficient information available at the start of the project and the right solution will almost certainly not already exist in big datasets or screening libraries. In other words, drug design is a learning - not a screening

- problem. This is true for both novel targets, where no work has been done before, and established targets, where new approaches must be devised that are distinct from existing efforts. As we start to explore novel chemical spaces, we are likely to be at the limit of predictive power, or the domain of applicability, for current models. Our systems and models are designed to learn and evolve, which, like nature, allows them to find optimised solutions to problems.

Integrated Experiment

Proprietary data and experimentation. We believe the development of proprietary data is critical to designing and developing the most differentiated medicines. We differentiate from others in the field by bringing experimentation in-house and integrating this with our technology platform. We have developed capabilities in assay development (generating our own proteins and cell reagents) and also sourcing tissue directly from patients. By bringing experimentation in-house we have been able to increase the underlying quality of the data we generate. High quality, reproducible experimental data is what drives our machine learning models.

Maximising information gain with efficient data collection. Bringing experimentation in-house also highlighted to us the importance of efficiency in data collection. Our teams leverage active learning when designing experiments. This enables us to efficiently select the right compounds to synthesise and test, to maximise the information gained that will drive the validity of our models. This is in keeping with our drug design principles; each experiment is important and is done to enable a learning system - rather than to generate big, mostly redundant, datasets for screening. The virtual platform this enables can then be integrated with precision experiments.

Evolution of integrated experiment. We believe the integration of AI with automation can bring significant efficiencies to the drug design process and accelerate information gain. The use of automation for chemical synthesis and experimentation has the potential to simultaneously reduce cycle times and multiplex biological evaluation. With automation there is a potential to explore more complex biology. For example, we can now investigate new assay types, and evaluate multiple experimental elements simultaneously (increasing the throughput of analysis). We have identified key aspects of our drug design and experiment process that should be automated and opened our state of the art automation facility, in Milton Park, Oxfordshire, in June 2023. We believe this will further help us scale the generation of valuable molecular intellectual property.

7

Exscientia plc | 2023 Annual Report and Financial Statements

Strategic Report

Closing the loop between the virtual and physical. We believe that we are the first company to have built an automation facility that has the potential to close the loop between AI-led drug design and experimentation. This facility has capabilities in compound management, automated chemical synthesis, automated biological screening, and in time we expect that it will enable us to produce proteins and develop DMPK assays . We have also integrated modules of AI generative design, active learning and AI retrosynthesis/chemical reaction design with the hardware. We also believe that we are the first company to develop software that can orchestrate synthesis and experimentation in the physical world with the computational precision design of compounds, driving the integration between the virtual world and the real world. This means we could learn more at a faster rate.

Precision Medicine Platform

The patient is the best model. Current model systems, such as outgrowth cell lines, do not feature the complex interplay of cells and environment necessary to model drug action prior to clinical studies. They are subject to culture adaptation, genetic drift and do not recapitulate the complexity of human disease. We utilise primary patient samples, as the most disease relevant model system, not just to model drug actions but also to identify next-generation targets. We deploy a wide variety of technologies and AI-driven data analysis techniques, such as custom deep learning algorithms for analysing images of primary cells after ex vivo drug perturbation. We further collect orthogonal multi- omics data including single cell transcriptomics, genomics, epigenetics and proteomics, enabling us to both quantify drug action and understand disease state. In the first-ever prospective interventional study of its kind, EXALT-1, our platform predicted which therapy was to be most effective for late- stage haematological cancer patients based on drug activity in their own tissue samples. EXALT-1 demonstrated the real-world patient selection capabilities of our platform with 54% of patients following the platform's recommendation demonstrating a clinical benefit of more than 1.3-fold enhanced progression-free survival (PFS) compared to their previous therapy.

8